Star3
7
Code
Code
Issues12
Pull Requests3
Discussion
Wiki
Analysis
Star3
7
Open
docs: align INTrainingUpdateGradGammaBeta 1076-case acceptance scope #9
raoliang_sacAdd on Sep 27
docs: align INTrainingUpdateGradGammaBeta 1076-case acceptance scope #9
Open
raoliang_sacAdd on Sep 27
Merged fromdocs/in-training-update-grad-gamma-beta-handoverintomain
Total: 270 files changed+24342-1724
Achangwei_deliverables/september/INTrainingUpdateGradGammaBeta/.gitattributes+2-0
@@ -0,0 +1,2 @@
1+# Preserve exact evidence bytes and portable SHA256 manifests.
2+** -text
Mchangwei_deliverables/september/INTrainingUpdateGradGammaBeta/00_check_findings.md+23-71
@@ -1,83 +1,35 @@
1-# INTrainingUpdateGradGammaBeta 转测交付件检查1+# INTrainingUpdateGradGammaBeta 交付审计
2 2 
3-扫描对象：`/workspace/cannbot-deliverables/INTrainingUpdateGradGammaBeta`3+## 离线审计范围
4 4 
5-扫描工具：5+- 标准精度表为原8列保持不变、扩展为12列2152行（增加rmse比、原始判定、指标类别、rmse缺失说明）；正式验收集1076个唯一用例、2152个输出，三方1076/1076 PASS；保留行原始指标不变。按用户授权仅排除范围清单中的两个GPU参考溢出用例；裁剪前原始失败记录完整保留，不代表原1078例三方全部通过。详见[独立验收范围清单](https://gitcode.com/zl_hw/myOpen/tree/49dc5703162b7e5ad146d32adbb1665984eb0242/handover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076)。指标分类为1718个numeric、432个ratio_not_applicable、2个empty_output；N/A不填0。
6+- Kernel200精度200/200 PASS，原SQLite审计最小GPU/NPU=0.540304不变，证据为归档:evidence/perf_final_20260927/。当前GEIR50例、100输出真实cross_check PASS；同次GPU与SQLite计时最小GPU/NPU=0.842386，50例均>0.1。check_invoke_path退出0。原始CST_OFF保留，不是TTK原生性能PASS。GEIR固定版本证据见[独立归档](https://gitcode.com/zl_hw/myOpen/tree/51cdd798f0809bffd40529f3223eef211ac00e1d/handover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928)。旧GEIR50 stat_rel_err及其0.575407比值仅为历史两方批次，不冒充当前三方。私有CANN补丁不等同正式canndev源码构建。
7+- 当前没有完整交付门禁或本地全量CI通过结论。
6 8 
7-- `changwei-op-dev/scripts/audit_deliverables.py`9+## 数值与环境限制
8-- `changwei-code-check/scripts/scan_handover.py`
9-- `ops-spec-gen/scripts/validate_spec.py --strict`
10 10 
11-## 结论11+- 正式验收binary两方与三方均1076/1076 PASS；历史1078条完整记录和两个GPU参考溢出FAIL保存在[独立验收范围清单](https://gitcode.com/zl_hw/myOpen/tree/49dc5703162b7e5ad146d32adbb1665984eb0242/handover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076)，范围裁剪不是修复GPU行为。
12+- 逐输出N/A应区分 ratio_not_applicable（可能涉及零值、非有限值或消去）、空输出、GPU FP32溢出；不能填0或计作三方PASS。
13+- TTK原始性能UNKNOWN/CST_OFF与SQLite时延审计分开陈述，不能将后处理时延标为TTK原生性能PASS。
14+- 私有CANN二进制补丁包含REG_OP屏蔽，不等同正式canndev源码构建验证；PR !82510已开不代表正式构建通过或已合入。
15+- 硬件实测仅Ascend950PR_9579，不外推为950DT板测。
12 16 
13-- 固定交付格式审计：**PASS**。17+## 人工判定
14-- PR 已 rebase 到检查时最新 `upstream/master`，并压缩为唯一算子提交
15- `1fe720b94fff1d9a9cbae12f2a9cc44a4a2e0936`；本地与远端分支一致。
16-- 配置与文档专项整改：**PASS**。已完成 REG_OP 唯一性、README 多产品差异、arch35 Host UT
17- SoC 隔离、op_graph InferDataType 归属、RegBase UB 指针和 CANN 9.0.0 编译兼容性检查。
18-- `spec.yaml` 11-stage 严格校验：**overall PASS**；stage 10 因双输入双输出无法映射为一次
19- `torch.sum` 调用而 SKIP，公式、shape、dtype、边界和 invariant 其余阶段通过。
20-- canndev 公共契约人工复核：**PASS**。FP32、4D NCHW/NHWC、两输入两输出、无属性、
21- `sum(axis=0, keepdims=True)` 和 GEIR 通路均一致。
22-- 当前存在 2 项 GEIR 通路覆盖度差异和 1 项低风险表格提示；它们不表示已测用例失败，也不构成当前已确认范围内的
23- canndev 兼容缺陷。
24 18 
25-## 自动扫描提示与人工判定19+- 仅支持 Kernel 直调与 GEIR；不支持 ACLNN、Torch 单算子及图模式、TensorFlow、ONNX 算子接口。Torch-A100 仅为第三方精度与性能参考，不是本算子接口。
20+- 07表另两处“不涉及”为本次范围外的融合Pass扩展和不支持的非连续Tensor；7处“不涉及”不改成虚假PASS。
21+- 运行模式已有 `--binary-only` / `-d=false -b=release` 记录；扫描器只查长参数字面值的提示不证明使用JIT。
22+- 质量PR !11235（head517193302、base720db44b7）净改动仅golden.py；CPU独立单测10/10 PASS，既有TTK代码路径CPU集成4/4 PASS；不改变历史设备证据版本归属。
23+- 历史检查通过项不自动沿用为当前全量放行结论；执行过程由原始日志追溯。
26 24 
27-### 1. invoke_path 数量低于通用门槛25+## 固定版本证据档案
28 26 
29-- 自动提示：GEIR 明细 29 条，低于工具要求的每通路 50 条。27+标准交付目录仅保留汇总说明与结果表，原始CSV/日志/SQLite及SHA清单位于[固定提交证据目录](https://gitcode.com/zl_hw/myOpen/tree/471e13bb05bf992586d6799ce310af4aff311089/changwei_deliverables/september/INTrainingUpdateGradGammaBeta/evidence)。文中的“归档:evidence/”以该固定版本为根，不是本地缺失目录。源版本myOpen `471e13bb05bf992586d6799ce310af4aff311089`，镜像版本cannbot-deliverables `cfbcd2efe374fcc8add51f0672776f627ecc0e7a`；均保留Git历史可达性。可按固定提交检出并在各证据子目录执行 `sha256sum -c SHA256SUMS`，精度审计脚本支持离线重放。
30-- 事实：当前提交的 GEIR 接受集为 29/29 PASS；原批次唯一 `GEInitialize failed` 为并发环境问题，
31- 对该项必要复验后 PASS。
32-- 判定：**覆盖度提示保留**。29 条足以支撑本轮当前实现验证结论，但若转测门禁硬性要求 ≥50 条，仍需新增
33- 至少 21 条 GEIR 用例后更新本表。
34 28 
35-### 2. GEIR 档没有同通路 A100 性能29+## GEIR通路证据限制
36 30 
37-- 自动提示：invoke_path 的 `xpu_device_us` 为空。31+历史GEIR50仅stat_rel_err，不能代表三方。当前GEIR50例、100输出真实cross_check PASS；同次GPU与SQLite计时最小GPU/NPU=0.842386，50例均>0.1。check_invoke_path退出0。原始CST_OFF保留，不是TTK原生性能PASS。当前证据采用PROF会话时间、进程PID和DBmtime联合绑定，并保存50例vendor运行时加载记录；私有补丁及正式源码构建限制仍保留。
38-- 事实：本轮 GEIR 档只验证 NPU release binary 与 CPU Golden，没有把 Kernel 档的 A100 数据冒充 GEIR
39- 同通路数据。独立三方精度 100/100 和 A5/A100 性能 200/200 已完成。
40-- 判定：**同通路证据差异保留**。当前功能结论有效；如果交付门禁要求“GEIR 同次三方精度和性能”，需要另行
41- 执行 GEIR+A100 同通路批次，不能拼接现有结果。
42 32 
43-### 3. 泛化规模已补齐33+## 当前GEIR原始证据
44 34 
45-- 补测：PR 当前 release binary 已执行 1000 条公共 4D FP32 Kernel 用例，NCHW/NHWC 各 500；35+[固定提交归档](https://gitcode.com/zl_hw/myOpen/tree/51cdd798f0809bffd40529f3223eef211ac00e1d/handover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928)位于标准交付目录外，246份源文件及原始SHA清单保持字节一致；含50SQLite和运行时vendor加载证据。myOpen证据提交`51cdd798f0809bffd40529f3223eef211ac00e1d`，cannbot镜像提交`ea0445df865cf2457e5ce0220983bff50b0cf083`。历史证据仍由前述旧固定commit访问，不被覆盖。
46- 1000/1000 精度、1000/1000 OOB、双输出 2000/2000 指标均 PASS。
47-- 累加：原 29 条全格式/边界定向集继续保留，覆盖 NCDHW/NDHWC/ND、rank4/5、空 Tensor、
48- 非对齐尾块、长归约、多 tile、强抵消、NaN/Inf 和有限溢出；两个集合身份无重复。
49-- 判定：**已闭环**。交付泛化明细更新为 1029 个唯一用例，满足通用 full 泛化门槛；补测未触发
50- 算子源码修改。
51- 
52-### 4. 自验证表“不涉及”较多
53- 
54-- 自动提示：`07_self_verification.xlsx` 中“不涉及”共 7 处。
55-- 人工复核：ACLNN、Torch 单算子、Torch 图、TensorFlow、ONNX 均无当前实现入口；Ascend 950 融合 Pass
56- 扩展不在本次范围；非连续 Tensor 没有 stride 接口且明确不支持。7 项均有源码和范围依据。
57-- 判定：**低风险提示已人工闭环**，不修改为虚假 PASS。
58- 
59-## 内容复核
60- 
61-| 检查项 | 结果 | 依据 |
62-|---|---|---|
63-| 7 项固定交付件 | PASS | 01–07 齐备 |
64-| 01 六节 / 02 九节结构 | PASS | 固定格式完整 |
65-| 四类测试结果 | PASS | 每阶段恰好一个 UTF-8 BOM CSV，标准表头且有明细 |
66-| 公共接口兼容 | PASS | 当前 proto/OpDef/Infer/Tiling/README 与 canndev 公共 REG_OP 对照 |
67-| 私有格式声明 | PASS | 明确不支持 `NDC1HWC0`，且不将其误列为公共兼容缺陷 |
68-| Golden 独立性 | PASS | CPU 高精度 Golden、Torch provider 与被测 Kernel 分离 |
69-| 三方精度 | PASS | 100/100 case、200/200 输出指标 |
70-| Kernel 泛化 | PASS | 公共集 1000/1000 + 定向集 29/29，合计 1029/1029 |
71-| 性能数据 | PASS | A5/A100 各 200/200 有效；逐例 `GPU/A5 > 0.1` 200/200 PASS，最小 0.554458 |
72-| 06/07 表格语义 | PASS | 每个 PASS/不涉及均与源码或证据对应 |
73- 
74-## 当前没有发现的问题
75- 
76-- 没有发现输入输出名称、顺序、dtype、公式或输出 shape 与 canndev 公共契约不一致。
77-- 没有发现交付件错误宣称 ACLNN 或私有 `NDC1HWC0` 支持。
78-- 没有将基础设施异常计为数值失败，也没有把无 profiling 的 `OK` 状态写成性能数值。
79-- 没有发现未推送的算子修改；当前 PR、本地和远端源分支均为
80- `1fe720b94fff1d9a9cbae12f2a9cc44a4a2e0936`。
81-- 数值证据在整改前提交 `6e6c9407...` 上采集。最终提交没有修改 Tiling、Kernel 计算体、Golden
82- 或用例输入；Kernel 入口只做文件名及 `\file` 注释同步，因此原精度/性能证据仍适用。最终提交另行通过
83- op_graph 4/4、op_host 38/38、无 SoC 聚合 18/18，以及 CANN 9.2/CANN 9.0.0 clean 定向构建。
Mchangwei_deliverables/september/INTrainingUpdateGradGammaBeta/01_requirement.md+12-19
@@ -1,15 +1,9 @@
1# INTrainingUpdateGradGammaBeta 需求文档1# INTrainingUpdateGradGammaBeta 需求文档
2 2 
3-> 交付对象：Cannbot 当前 Ascend 950 实现，ops-nn PR3+> 交付对象：已通过 ops-nn [PR !11067](https://gitcode.com/cann/ops-nn/merge_requests/11067)
4-> [!10488](https://gitcode.com/cann/ops-nn/merge_requests/10488)，源码提交4+> 合入 master 的 Cannbot Ascend 950 实现，合入提交 `40250c5c13436b40da9aeb5ca9bf75beab7a6058`。
5-> `1fe720b94fff1d9a9cbae12f2a9cc44a4a2e0936`。5+> Golden 质量加固 [PR !11235](https://gitcode.com/cann/ops-nn/merge_requests/11235)
6- 6+> head `517193302`、base `720db44b7`，仅涉及测试资产。PR !10488 / `1fe720b9` 是历史交付基线。
7-## 修订记录
8- 
9-| 版本 | 修订内容 | 修订时间 |
10-|---|---|---|
11-| v1.0 | 按当前 PR 源码与最终验证证据整理转测需求 | 当前版 |
12-| v1.1 | 同步 rebase 后单提交及配置/文档专项整改结果 | 当前版 |
13 7 
14## 1. 功能概述8## 1. 功能概述
15 9 
@@ -107,26 +101,25 @@ canndev 存在 `SingleInstanceNormGradFusion` 历史 Pass，但在 Ascend 950
107 101 
108### 6.1 功能与精度102### 6.1 功能与精度
109 103 
110-- op_graph、Host/InferShape/Tiling UT 与 Golden UT 全部通过。104+- op_graph、Host/InferShape/Tiling UT 与 Golden UT 应按对应源码版本验证，以具名证据为准。
111-- Kernel release binary、GEIR、动态 OOM 插桩 safety 全部通过。105+- Kernel release binary、GEIR、动态 OOM 插桩 safety 按各自版本和用例集合验证，不用单一路径结果替代其他路径。
112- Kernel 泛化至少覆盖 1000 条公共 4D FP32 NCHW/NHWC 用例；原定向回归继续保留，不用新结果覆盖旧证据。106- Kernel 泛化至少覆盖 1000 条公共 4D FP32 NCHW/NHWC 用例；原定向回归继续保留，不用新结果覆盖旧证据。
113- 三方精度使用同一输入完成 Ascend 950 release binary、CPU 高精度 Golden 与 Torch-A100107- 三方精度使用同一输入完成 Ascend 950 release binary、CPU 高精度 Golden 与 Torch-A100
114 `cross_check`；两个输出分别判定，不允许用整体平均掩盖一路失败。108 `cross_check`；两个输出分别判定，不允许用整体平均掩盖一路失败。
115-- 三方精度接受集为 100 条公共格式用例，不因性能扩容重复执行 ACLNN 或无关前端测试。109+- 正式验收集1076个唯一用例、2152个输出，三方1076/1076 PASS；保留行原始指标不变。按用户授权仅排除范围清单中的两个GPU参考溢出用例；裁剪前原始失败记录完整保留，不代表原1078例三方全部通过。详见[独立验收范围清单](https://gitcode.com/zl_hw/myOpen/tree/49dc5703162b7e5ad146d32adbb1665984eb0242/handover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076)。指标分类为1718个numeric、432个ratio_not_applicable、2个empty_output；N/A不填0。不涉及ACLNN/Torch算子接口。
116- FP32 Golden 判据采用仓内 `stat_rel_err`，阈值 `1e-4`；NaN/Inf 另做类别一致性判断。110- FP32 Golden 判据采用仓内 `stat_rel_err`，阈值 `1e-4`；NaN/Inf 另做类别一致性判断。
117 111 
118### 6.2 性能112### 6.2 性能
119 113 
120-- 性能在精度通过后采集 200 条，A5 与 A100 使用同一批 shape；A100 基线为两次原生114+- 性能证据范围为 Kernel 200 条与 GEIR 50 条，NPU 与 A100 使用对应相同 shape；A100 基线为两次原生
121 `torch.sum(dim=0, keepdim=True)` 的 eager/composite 实现，因 PyTorch 无同语义 fused/native 单接口。115 `torch.sum(dim=0, keepdim=True)` 的 eager/composite 实现，因 PyTorch 无同语义 fused/native 单接口。
122-- A5 使用 TTK `bin_perf_us`；A100 使用 Nsight 统计每个 NVTX case 范围内两个 CUDA116+- Kernel200精度200/200 PASS，原SQLite审计最小GPU/NPU=0.540304不变，证据为归档:evidence/perf_final_20260927/。当前GEIR50例、100输出真实cross_check PASS；同次GPU与SQLite计时最小GPU/NPU=0.842386，50例均>0.1。check_invoke_path退出0。原始CST_OFF保留，不是TTK原生性能PASS。GEIR固定版本证据见[独立归档](https://gitcode.com/zl_hw/myOpen/tree/51cdd798f0809bffd40529f3223eef211ac00e1d/handover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928)。旧GEIR50 stat_rel_err及其0.575407比值仅为历史两方批次，不冒充当前三方。私有CANN补丁不等同正式canndev源码构建。
123- reduction kernel 的总耗时，不计算 SSH/HTTP 传输时间。117+- 每条用例要求 `GPU device_us / NPU kernel us > 0.1`；除此底线外性能只报告，
124-- 每条用例要求 `GPU device_us / A5 release-binary us > 0.1`；除此底线外性能只报告，
125 不以性能数据替代功能与精度结论。118 不以性能数据替代功能与精度结论。
126 119 
127### 6.3 交付证据120### 6.3 交付证据
128 121 
129-- 精度/性能 CSV 保留其实际采集基线 `6e6c9407...`；最终 PR 提交 `1fe720b9...` 必须提供差异证明，122+- 历史精度/性能 CSV 的采集基线 `6e6c9407...` 与历史 PR 提交 `1fe720b9...` 的迁移证明，仅适用于该历史版本的 Tiling、Kernel、Golden、输入及配置回归，不作为当前 PR 的新增设备验证。
130- 确认 Tiling、Kernel 计算体、Golden 与用例输入未变，并完成配置归属相关的最小必要回归。123+- 当前质量 PR !11235 与已有验证基线 `17da22f71` 分开标识；保存原始 CSV、逐输出状态和来源哈希。只有完成来源对账的结果才能纳入交付结论。
131- TTK 每次执行显式 `--pc=1`；构建并发不超过 8。124- TTK 每次执行显式 `--pc=1`；构建并发不超过 8。
132- 环境失败与数值失败分开记录，只允许对确认的环境失败做必要复验，并保留原始记录。125- 环境失败与数值失败分开记录，只允许对确认的环境失败做必要复验，并保留原始记录。
Mchangwei_deliverables/september/INTrainingUpdateGradGammaBeta/02_design.md+11-24
@@ -1,15 +1,10 @@
1# INTrainingUpdateGradGammaBeta arch35 设计文档1# INTrainingUpdateGradGammaBeta arch35 设计文档
2 2 
3-> 对应 ops-nn PR !10488，源码提交3+> 实现已通过 ops-nn PR !11067 合入 master，合入提交 `40250c5c13436b40da9aeb5ca9bf75beab7a6058`。
4-> `1fe720b94fff1d9a9cbae12f2a9cc44a4a2e0936`。本文描述当前已验证实现，不采用早期 Cannbot4+> Golden 质量加固 PR !11235 为 head `517193302` / base `720db44b7`，仅涉及测试资产。
5-> 资产中已经撤销的 ACLNN 或旧多 TilingKey 方案。5+> PR !10488 / `1fe720b9` 属于历史设计及验证基线；本算子不提供 ACLNN，也不采用旧多 TilingKey 方案。
6 6 
7-## 修订记录7+> 设计保留历史基线 `1fe720b9` 的来源；已有设备证据基于 `17da22f71`，不自动等同当前 PR head 全量验证。
8- 
9-| 版本 | 修订内容 | 修订时间 |
10-|---|---|---|
11-| v1.0 | 按当前单提交源码和最终验证结果整理 | 当前版 |
12-| v1.1 | 同步配置归属整改、CANN 9.0.0 构建与 rebase 后提交 | 当前版 |
13 8 
14## 1. 范式与模板选型9## 1. 范式与模板选型
15 10 
@@ -145,7 +140,7 @@ struct INTrainingUpdateGradGammaBetaTilingData {
145| canndev 公共兼容面 | 4D NCHW、NHWC |140| canndev 公共兼容面 | 4D NCHW、NHWC |
146| 950 扩展 | 5D NCDHW、NDHWC；ND rank 4/5；空 Tensor |141| 950 扩展 | 5D NCDHW、NDHWC；ND rank 4/5；空 Tensor |
147| 动态能力 | dynamic compile static、dynamic shape、dynamic rank |142| 动态能力 | dynamic compile static、dynamic shape、dynamic rank |
148-| 通路 | GE IR 节点；无 ACLNN/Torch/TF/ONNX 单算子接口 |143+| 通路 | 仅支持 Kernel 直调与 GEIR（GE IR 节点）；不支持 ACLNN/Torch/TF/ONNX 算子接口；Torch-A100 仅为第三方精度与性能参考 |
149| 不支持 | `NDC1HWC0`、`NC1HWC0`、非连续 tensor、输入输出 alias |144| 不支持 | `NDC1HWC0`、`NC1HWC0`、非连续 tensor、输入输出 alias |
150 145 
151## 7. 交付物146## 7. 交付物
@@ -181,17 +176,9 @@ norm/in_training_update_grad_gamma_beta/
181 176 
182## 9. 验证设计与结果177## 9. 验证设计与结果
183 178 
184-- 最终提交定向回归：op_graph 4/4、Ascend 950 op_host 38/38、无 SoC 聚合 op_host 18/18 PASS。179+- Kernel/Golden 使用 stat_rel_err，FP32阈值1e-4；三方使用NPU / CPU FP64 Golden / Torch-A100 cross_check，两个输出独立判定。
185-- CANN 9.2 与独立安装的 CANN 9.0.0 均完成 `--ccache=off` clean 定向构建，生成 5 组180+- 正式验收Binary两方1076/1076 PASS；正式验收集1076个唯一用例、2152个输出，三方1076/1076 PASS；保留行原始指标不变。按用户授权仅排除范围清单中的两个GPU参考溢出用例；裁剪前原始失败记录完整保留，不代表原1078例三方全部通过。详见[独立验收范围清单](https://gitcode.com/zl_hw/myOpen/tree/49dc5703162b7e5ad146d32adbb1665984eb0242/handover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076)。指标分类为1718个numeric、432个ratio_not_applicable、2个empty_output；N/A不填0。
186- Kernel object/JSON；构建并发均为 `-j8`。181+- Golden CPU单测10/10、TTK代码路径CPU集成4/4 PASS。历史1fe720b9的op_graph/Host UT、CANN定向构建和safety结果只适用于对应历史基线，不作为当前head新增验证。
187-- Golden pytest：7/7 PASS。182+- GEIR定向29/29两方PASS且双推导命中，见 `归档:evidence/geir_20260926/`；此前50例单独保存。私有CANN字符串补丁同时屏蔽REG_OP，不等同正式canndev源码构建验证，PR !82510不构成正式构建通过证明。
188-- Kernel release binary：新增公共 NCHW/NHWC 泛化 1000/1000 PASS；原全格式/边界定向集183+- Kernel200精度200/200 PASS，原SQLite审计最小GPU/NPU=0.540304不变，证据为归档:evidence/perf_final_20260927/。当前GEIR50例、100输出真实cross_check PASS；同次GPU与SQLite计时最小GPU/NPU=0.842386，50例均>0.1。check_invoke_path退出0。原始CST_OFF保留，不是TTK原生性能PASS。GEIR固定版本证据见[独立归档](https://gitcode.com/zl_hw/myOpen/tree/51cdd798f0809bffd40529f3223eef211ac00e1d/handover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928)。旧GEIR50 stat_rel_err及其0.575407比值仅为历史两方批次，不冒充当前三方。私有CANN补丁不等同正式canndev源码构建。
189- 29/29 PASS，合计 1029 个唯一用例，精度和 OOB 全通过。184+- 历史标准通路29例与性能200例保存在最终性能证据previous_performance.csv与previous_invoke_path.csv中，不与当前标准表混算。范围与限制见00_check_findings.md、00_check_findings.md。
190-- GEIR：首批 28/29；唯一项为并发环境下 `GEInitialize failed`，必要复验 1/1 PASS，合并 29/29。
191-- 动态 OOM 插桩 safety：29/29 PASS。
192-- 三方扩展集：100/100 case、两个输出 200/200 `cross_check` PASS；10 条 GPU 网关心跳假阴性仅做必要复验。
193-- 100 条三方精度保持不变；性能使用 200 条公共矩阵，复用已验证的 100 条并只补其精确补集 100 条。A5 200/200、A100 200/200 有效，逐例 `GPU/A5 > 0.1` 200/200 PASS；比值最小 0.554458、p50 3.205499。
194-- A100 无同语义 fused/native PyTorch 单接口，因此使用未编译的 eager/composite 两次 `torch.sum`；Nsight 按 NVTX case 范围累加 CUDA kernel 时间，每条每次测量恰好两个 reduction kernel。
195-- 数值证据对应整改前提交 `6e6c9407...`。最终提交 `1fe720b9...` 只调整注册/UT 隔离、
196- InferDataType 归属、文档与 Kernel 入口文件名；Tiling、Kernel 计算体、Golden 和测试输入均未改，
197- 因此没有再次执行精度/性能全量集，改用上述定向 UT 与双版本 CANN clean 构建闭环。
Mchangwei_deliverables/september/INTrainingUpdateGradGammaBeta/03_spec.yaml+2-1
@@ -1,7 +1,8 @@
1# ============================================================1# ============================================================
2# in_training_update_grad_gamma_beta — L0 数学约束 spec2# in_training_update_grad_gamma_beta — L0 数学约束 spec
3# 覆盖范式: Reduction3# 覆盖范式: Reduction
4-# 交付真值上游: 01_requirement.md + ops-nn PR !10488 当前源码4+# 交付契约来源: 01_requirement.md + ops-nn PR !11067 已合入实现
5+# PR !10488 / 1fe720b9 为历史基线；质量 PR !11235 仅改测试资产，不改本数学契约
5# ============================================================6# ============================================================
6schema_version: 17schema_version: 1
7 8 
Mchangwei_deliverables/september/INTrainingUpdateGradGammaBeta/04_golden/GOLDEN_REF.md+32-42
@@ -1,64 +1,54 @@
1-# Golden 引用1+# Golden 引用（质量 PR !11235）
2 2 
3-## 代码位置3+## 代码位置与版本
4 4 
5-- 生产仓 Golden：`norm/in_training_update_grad_gamma_beta/tests/assets/golden.py`。5+- 生产仓 Golden：`ops-nn/norm/in_training_update_grad_gamma_beta/tests/assets/golden.py`；质量 PR !11235 head `517193302`、base `720db44b7`。已有设备证据基于分支 `quality/in-training-update-grad-gamma-beta-golden` 的 `17da22f71` / 上游 `65f130e2`，不改标为新 head 的全量验证。
6-- 数值证据采集提交：`6e6c9407968a3af2fb0812276b219cb6ab07ed61`。6+- 历史迭代提交：`dadb698a4`（高精度 Golden 重写）→ `17da22f71`（移植 changwei 9 条 DFX 输入注入 + torch 延迟导入）；当前 PR 已重基至 `720db44b7`，head `517193302`，Golden 文件 SHA256 未变。
7-- 最终 PR 提交：`1fe720b94fff1d9a9cbae12f2a9cc44a4a2e0936`；Golden 文件与采集提交一致。7+- Golden 文件 SHA256：`69afaa6f90220dbf26e2d3fd964a8489915a9ea2921bfe13a84a02c3e986559d`。
8-- 三方精度验证快照：`validation-100-moderate-20260921/plugins/golden.py`。8+- 上述SHA256标识当前Golden；每份验证证据按其原始版本记录归属，不混算历史不同Golden的结果。
9-- 性能竞品脚本快照：`kernel1000-performance200-20260922/gpu_performance_200.py`。
10-- Golden 文件 SHA256：`07e60cca8998a95dad85d8f89eaa6fcea740e07943cd54cbc2f6607674256a07`。
11 9 
12## 数学语义10## 数学语义
13 11 
14```python12```python
15pd_gamma = sum(res_gamma, axis=0, keepdims=True)13pd_gamma = sum(res_gamma, axis=0, keepdims=True)
16-pd_beta = sum(res_beta, axis=0, keepdims=True)14+pd_beta = sum(res_beta, axis=0, keepdims=True)
17```15```
18 16 
19-`InTrainingUpdateGradGammaBetaSpec.golden(res_gamma, res_beta, **kwargs)` 返回17+`InTrainingUpdateGradGammaBetaSpec.golden(res_gamma, res_beta, **kwargs)` 返回 `[pd_gamma, pd_beta]`。两路独立生成和判定，不复用被测 Ascend C Kernel 的实现或中间结果。
20-`[pd_gamma, pd_beta]`。两路独立生成和判定，不复用被测 Ascend C Kernel 的实现或中间结果。
21 18 
22## CPU Golden19## CPU Golden
23 20 
24-- 输入先按 FP32 解释，通常使用 NumPy FP64 沿 dim0 累加后转回 FP32。21+- 输入**不再强制转 FP32**：CPU 真值使用 FP64 并返回 FP64（保留 TTK Promote 精度）。
25-- 对误差界显示为高风险的有限列，使用 Python `math.fsum` 恢复高精度求和，避免普通 FP64 顺序累加22+- NumPy FP64 沿 dim0 累加；对误差界显示为高风险的有限列（强抵消）用 Python `math.fsum` 恢复，避免普通 FP64 顺序累加成为错误参考。
26- 在极端抵消场景下成为错误参考。23+- NaN、正 Inf、负 Inf 和有限值溢出按最终 FP32 类别保留；`N=0` 生成全 0 输出；非归约维为 0 时输出为空 Tensor。
27-- NaN、正 Inf、负 Inf 和有限值溢出按最终 FP32 类别保留。24+- 输入注入 `customize_inputs` 支持两套命名：cannbot 定向集（`_multiscale_*`/`overflow_boundary_*`/`scale_boundary_neighbors`/`extreme_*` 等）与 changwei 9 条 DFX（`_dfx_zero/nan/single_inf/mixed_inf/finite_pattern/subnormal/cancellation/gamma_zero_beta_pattern/finite_overflow`，基线17da22f71 从历史分支移植，注入自检 9/9）。
28-- `N=0` 生成全 0 输出；非归约维为 0 时输出为空 Tensor。
29 25 
30## Torch-A100 第三方参考26## Torch-A100 第三方参考
31 27 
32-- 三方精度使用独立 `_TorchReference`：Torch FP64 归约，高风险列使用 Decimal 恢复；不调用 CPU28+- `third_party = {"torch": _TorchReference}`：**两个原生 `torch.sum(dim=0, keepdim=True)`，保持输入 dtype/设备**；去掉 FP64/CPU Decimal 回退（相对上一迭代的变更，后续三方与性能须绑定该参考）。
33- Golden 的 `math.fsum` 路径。29+- `_TorchReference` 仅用于远端GPU参考计算，`_sum` 使用延迟导入；该参考不提供本算子的Torch调用接口。
34-- TTK 固定使用 `--provider=torch --compare=cross_check --pc=1`，同一输入下同时记录 NPU、CPU30+- TTK `--compare cross_check`（L1）；历史端点gpu-20101，当前采集端点gpu-20102，按各批来源标识。正式验收集1076个唯一用例、2152个输出，三方1076/1076 PASS；保留行原始指标不变。按用户授权仅排除范围清单中的两个GPU参考溢出用例；裁剪前原始失败记录完整保留，不代表原1078例三方全部通过。详见[独立验收范围清单](https://gitcode.com/zl_hw/myOpen/tree/49dc5703162b7e5ad146d32adbb1665984eb0242/handover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076)。指标分类为1718个numeric、432个ratio_not_applicable、2个empty_output；N/A不填0。
35- Golden 和 Torch provider 结果。
36-- 性能参考与精度参考分离：性能插件在 A100 上执行两次原生
37- `torch.sum(dim=0, keepdim=True)`。PyTorch 无同语义 fused/native 单接口，因此明确标记为
38- eager/composite 基线，`torch.compile` 关闭。
39-- A100 性能使用 ops-competitor-profiling/Nsight：输入在 GPU 主机本地生成，预热 3 次、
40- 测量 5 次，对每个 NVTX case 范围内恰好两个 CUDA reduction kernel 的时间求和，
41- 不把 SSH/HTTP 传输时间计入设备耗时。
42 31 
43## 比对判据32## 比对判据
44 33 
45-- 常规 Kernel/Golden 判据：`stat_rel_err`，FP32 threshold=`1e-4`。34+- 常规 Kernel/Golden 判据：`stat_rel_err`，FP32 threshold=`1e-4`（官方标准，无自定义 compare/pre_compare）。
46-- 三方用例：两个输出的 NPU 和 Torch 指标分别对 Golden 执行 `cross_check`；所有实际输出指标均通过，35+- 三方用例：NPU / CPU Golden / Torch-A100 三路 `cross_check`；两个输出独立判定。
47- case 才计为 PASS。
48- NaN、+Inf、-Inf 按类别和位置一致性判断，不用有限值相对误差掩盖特殊值错误。36- NaN、+Inf、-Inf 按类别和位置一致性判断，不用有限值相对误差掩盖特殊值错误。
49-- 性能状态只在对应精度/OOB 已通过后接受；200 条性能逐例要求 `GPU/A5 > 0.1`，除此之外不设置37+- 已知边界：`overflow_boundary_equal/below`（和距 max_f32 约 0.6 ULP）上 GPU torch.sum FP32 累加溢出为 inf，NPU 与 FP32(Golden) 位级一致——定性为 GPU 参考实现差异，留档不放宽阈值。
50- 额外性能合入门槛。
51 38 
52-## 最终证据39+## 已有验证证据（基线17da22f71）
53 40 
54-最终提交没有修改 Tiling、Kernel 计算体、Golden 或测试输入；Kernel 入口只做文件名及 `\file`41+- 正式验收两方（NPU vs CPU Golden）：1076/1076 PASS，OOB全OK；裁剪前1078条两方完整记录见[独立验收范围清单](https://gitcode.com/zl_hw/myOpen/tree/49dc5703162b7e5ad146d32adbb1665984eb0242/handover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076)。
55-注释同步。因此以下数值证据继续有效，且不伪装为在最终提交上重新采集。42+- 正式验收三方（cross_check）：1076/1076 PASS；范围裁剪及原始失败记录见[独立验收范围清单](https://gitcode.com/zl_hw/myOpen/tree/49dc5703162b7e5ad146d32adbb1665984eb0242/handover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076)。
43+- CPU 回归：`tests/ut/test_golden.py` 10/10、`quality_hardening/golden_audit/test_ttk_integration.py` 4/4。
44+- 已有状态汇总见 `../05_test_result/generalization/` 与 `../05_test_result/precision/`；原始指标与批次来源应按审计映射追溯。
56 45 
57-- 29 条全格式三方回归：29/29 case、58/58 输出指标 PASS。46+## 历史数值证据（上一迭代，基线 `1fe720b9`，仅追溯）
58-- 公共 Kernel 泛化：1000/1000 case、2000/2000 输出指标、1000/1000 OOB PASS；与原 29 条定向集47+ 
59- 累加后为 1029 个唯一 Kernel 用例。48+上一迭代的 1029 条泛化 / 29 条 GEIR / 200 条性能数据基于旧 Golden（SHA256 `07e60cca...`，含 Torch FP64+Decimal 三方参考）与旧基线采集；三方参考口径已变更，旧三方/性能结果**不可**与本次新口径混算。
60-- 最新 100 条中等 shape 公共格式扩展集：100/100 case、200/200 输出指标 PASS。49+ 
61-- 初始批次 100 条中的 10 条失败只因共享 GPU 网关心跳假阴性、未取得 provider 结果；仅复验这 10 条后50+## GEIR 增量验证
62- 10/10 PASS。初始批次和复验记录均保留，未将环境失败描述为数值失败。51+ 
63-- 性能矩阵：A5/A100 各 200/200 有效，逐例 `GPU/A5 > 0.1` 全部通过；最小值52+Golden SHA256保持上述版本；CPU10/10与TTK代码路径CPU集成4/4 PASS。GEIR 新增 29 个不同用例两方精度 PASS，全部命中新 InferShape/InferDataType；此前 50 PASS 单独保留。该定向GEIR批次为在线编译/执行，非三方、非 binary-only、非性能。证据见 `归档:evidence/geir_20260926/`。
64- 0.554458、p50 3.205499。A5 在 178 条上更快，A100 在 22 条上更快。53+ 
54+Kernel200精度200/200 PASS，原SQLite审计最小GPU/NPU=0.540304不变，证据为归档:evidence/perf_final_20260927/。当前GEIR50例、100输出真实cross_check PASS；同次GPU与SQLite计时最小GPU/NPU=0.842386，50例均>0.1。check_invoke_path退出0。原始CST_OFF保留，不是TTK原生性能PASS。GEIR固定版本证据见[独立归档](https://gitcode.com/zl_hw/myOpen/tree/51cdd798f0809bffd40529f3223eef211ac00e1d/handover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928)。旧GEIR50 stat_rel_err及其0.575407比值仅为历史两方批次，不冒充当前三方。私有CANN补丁不等同正式canndev源码构建。
Mchangwei_deliverables/september/INTrainingUpdateGradGammaBeta/05_test_result/invoke_path/INTrainingUpdateGradGammaBeta_invoke_path.csv+62-35
@@ -1,38 +1,65 @@
1项,内容1项,内容
2-支持通路范围,kernel=支持；GEIR=支持；aclnn/torch/tf/onnx=不涉及。自动融合 Pass 扩展不在本次范围。2+支持通路范围,仅支持Kernel直调与GEIR；不支持ACLNN/Torch算子接口。Torch-A100仅第三方参考；Kernel结果见generalization/precision/performance。
3-判据,GEIR release binary 双输出精度通过；用例与当前契约一致。3+判据,同次GEIR双输出cross_check L1：mare比≤5、mere比≤1.5、rmse比≤1.5；原始TTK判定及小值域规则保留在逐输出证据。
4-覆盖与裁剪,29 条覆盖公共 NCHW/NHWC 和 Ascend950 扩展 NCDHW/NDHWC/ND、空 Tensor及数值边界。4+覆盖与裁剪,原50例相同输入，五个accepted十例批次；50/50用例、100/100输出三方PASS，两个污染attempt排除。
5-结论,原批次 28/29 PASS；唯一异常为同机并发期间 GEInitialize failed，必要复验该 1 条 PASS，合并接受集 29/29 PASS。5+结论,连通/真实三方精度/GPU计时均50/50；SQLite时延审计50/50，最小GPU/NPU=0.842386，逐例>0.1。
6-原始证据,geir-bin-29-01.csv + geir-bin-retry-ncdhw-full-01.csv；本表用复验 PASS 行替换原环境失败行。6+取数方法,GPU为同次torch profiler device_us；NPU取本轮SQLite task_time.duration_time，ns/1000，每例1条kernel task。
7-复现口径,TTK geir，release binary，显式 --pc=1。7+新鲜度与来源,按PROF会话时间+GEIR进程PID/用例映射+DBmtime联合绑定，避免仅mtime误认旧会话；50份DB SHA及运行时vendor库加载可复核。
8+TTK原始性能,cst_perf_us=CST_OFF 50/50；SQLite后处理计时不是TTK原生性能PASS。
9+折算比,N/A：无可靠硬件折算因子，不以raw G/N冒充folded_G/N。
10+复现方法,强制TTK技能入口geir，--pc=1 --compare cross_check --xpu-perf --task-prof=true；原始输入/脚本/配置及同次日志见固定提交证据。
11+证据,https://gitcode.com/zl_hw/myOpen/tree/51cdd798f0809bffd40529f3223eef211ac00e1d/handover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928
12+环境限制,仅Ascend950PR_9579实测；私有CANN注册字符串补丁包含REG_OP屏蔽，不等同正式canndev源码构建；不宣称vendor库由Golden-only提交整包重建。
13+历史范围,此前GEIR50 stat_rel_err属于独立历史两方批次，不与当前cross_check结果混算。
8 14 
9通路,testcase,dtype,shape,attributes,precision_status,xpu_status,xpu_device_us,npu_perf_us,raw_G/N,folded_G/N,结论15通路,testcase,dtype,shape,attributes,precision_status,xpu_status,xpu_device_us,npu_perf_us,raw_G/N,folded_G/N,结论
10-geir,in_training_update_grad_gamma_beta_fp32_nchw_base,float32,"NCHW (4, 64, 3, 5)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS16+geir,in_training_update_grad_gamma_beta_gen0000_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1, 1, 1, 1], [1, 1, 1, 1]]",{},PASS(双输出cross_check),PASS(GPU参考),8.52,3.124,2.727273,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
11-geir,in_training_update_grad_gamma_beta_fp32_nhwc_base,float32,"NHWC (4, 3, 5, 64)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS17+geir,in_training_update_grad_gamma_beta_gen0020_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 1, 1, 648], [2, 1, 1, 648]]",{},PASS(双输出cross_check),PASS(GPU参考),14.2,4.57,3.107221,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
12-geir,in_training_update_grad_gamma_beta_fp32_ncdhw_base,float32,"NCDHW (2, 16, 3, 4, 5)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS18+geir,in_training_update_grad_gamma_beta_gen0041_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[127, 5, 15, 1], [127, 5, 15, 1]]",{},PASS(双输出cross_check),PASS(GPU参考),32.0,8.276,3.866602,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
13-geir,in_training_update_grad_gamma_beta_fp32_ndhwc_base,float32,"NDHWC (2, 3, 4, 5, 16)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS19+geir,in_training_update_grad_gamma_beta_gen0061_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 7, 5, 7], [16, 7, 5, 7]]",{},PASS(双输出cross_check),PASS(GPU参考),12.8,5.754,2.224539,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
14-geir,in_training_update_grad_gamma_beta_fp32_nd_alias,float32,"ND (4, 64, 3, 5)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS20+geir,in_training_update_grad_gamma_beta_gen0081_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[31, 7, 3, 127], [31, 7, 3, 127]]",{},PASS(双输出cross_check),PASS(GPU参考),20.2,6.29,3.211447,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
15-geir,in_training_update_grad_gamma_beta_fp32_single_n,float32,"NCHW (1, 64, 3, 5)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS21+geir,in_training_update_grad_gamma_beta_gen0102_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[8, 5, 3, 65], [8, 5, 3, 65]]",{},PASS(双输出cross_check),PASS(GPU参考),15.0,5.48,2.737226,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
16-geir,in_training_update_grad_gamma_beta_fp32_large_n_group,float32,"NCHW (100000, 8, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS22+geir,in_training_update_grad_gamma_beta_gen0122_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[5, 3, 32, 16], [5, 3, 32, 16]]",{},PASS(双输出cross_check),PASS(GPU参考),15.3,6.12,2.5,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
17-geir,in_training_update_grad_gamma_beta_fp32_wide_channels,float32,"NCHW (8, 1024, 7, 7)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS23+geir,in_training_update_grad_gamma_beta_gen0142_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[16, 7, 32, 648], [16, 7, 32, 648]]",{},PASS(双输出cross_check),PASS(GPU参考),25.9,30.746,0.842386,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
18-geir,in_training_update_grad_gamma_beta_fp32_reduce_empty_n0,float32,"NCHW (0, 64, 3, 5)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS24+geir,in_training_update_grad_gamma_beta_gen0162_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[7, 7, 32, 4], [7, 7, 32, 4]]",{},PASS(双输出cross_check),PASS(GPU参考),15.4,5.598,2.750982,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
19-geir,in_training_update_grad_gamma_beta_fp32_zero_a_empty,float32,"NCHW (4, 0, 3, 5)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS25+geir,in_training_update_grad_gamma_beta_gen0184_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[127, 7, 1, 5], [127, 7, 1, 5]]",{},PASS(双输出cross_check),PASS(GPU参考),31.7,8.564,3.701541,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
20-geir,in_training_update_grad_gamma_beta_fp32_nchw_unaligned_tail,float32,"NCHW (7, 13, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS26+geir,in_training_update_grad_gamma_beta_gen0204_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1, 31, 8, 65], [1, 31, 8, 65]]",{},PASS(双输出cross_check),PASS(GPU参考),14.5,5.536,2.61922,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
21-geir,in_training_update_grad_gamma_beta_fp32_ndhwc_unaligned_tail,float32,"NDHWC (3, 1, 1, 1, 17)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS27+geir,in_training_update_grad_gamma_beta_gen0224_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 1, 16, 648], [2, 1, 16, 648]]",{},PASS(双输出cross_check),PASS(GPU参考),14.6,5.224,2.794793,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
22-geir,in_training_update_grad_gamma_beta_fp32_extreme_cancellation,float32,"NCHW (4, 13, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS28+geir,in_training_update_grad_gamma_beta_gen0245_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[63, 16, 5, 2], [63, 16, 5, 2]]",{},PASS(双输出cross_check),PASS(GPU参考),22.9,6.274,3.649984,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
23-geir,in_training_update_grad_gamma_beta_fp32_extreme_overflow,float32,"NCHW (2, 13, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS29+geir,in_training_update_grad_gamma_beta_gen0265_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 4, 2, 4], [3, 4, 2, 4]]",{},PASS(双输出cross_check),PASS(GPU参考),11.8,4.752,2.483165,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
24-geir,in_training_update_grad_gamma_beta_fp32_multi_tile_tail,float32,"NCHW (2, 800003, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS30+geir,in_training_update_grad_gamma_beta_gen0285_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[64, 1, 31, 5], [64, 1, 31, 5]]",{},PASS(双输出cross_check),PASS(GPU参考),21.6,6.84,3.157895,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
25-geir,in_training_update_grad_gamma_beta_fp32_extreme_nonfinite,float32,"NCHW (4, 8, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS31+geir,in_training_update_grad_gamma_beta_gen0306_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[4, 1, 3, 1], [4, 1, 3, 1]]",{},PASS(双输出cross_check),PASS(GPU参考),10.6,4.166,2.544407,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
26-geir,in_training_update_grad_gamma_beta_nchw_overflow_boundary_equal,float32,"NCHW (10000, 13, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS32+geir,in_training_update_grad_gamma_beta_gen0326_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[3, 31, 5, 65], [3, 31, 5, 65]]",{},PASS(双输出cross_check),PASS(GPU参考),14.3,5.79,2.469775,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
27-geir,in_training_update_grad_gamma_beta_nhwc_overflow_boundary_below,float32,"NHWC (10000, 1, 1, 13)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS33+geir,in_training_update_grad_gamma_beta_gen0346_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[31, 2, 32, 31], [31, 2, 32, 31]]",{},PASS(双输出cross_check),PASS(GPU参考),21.0,6.3,3.333333,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
28-geir,in_training_update_grad_gamma_beta_ncdhw_scale_boundary_neighbors,float32,"NCDHW (10000, 13, 1, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS34+geir,in_training_update_grad_gamma_beta_gen0367_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[8, 16, 7, 3], [8, 16, 7, 3]]",{},PASS(双输出cross_check),PASS(GPU参考),12.2,5.398,2.260096,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
29-geir,in_training_update_grad_gamma_beta_ndhwc_scale_boundary_neighbors,float32,"NDHWC (8193, 1, 1, 1, 17)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS35+geir,in_training_update_grad_gamma_beta_gen0387_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[257, 16, 16, 15], [257, 16, 16, 15]]",{},PASS(双输出cross_check),PASS(GPU参考),30.2,22.368,1.350143,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
30-geir,in_training_update_grad_gamma_beta_nd_cancellation_orders,float32,"ND (4, 24, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS36+geir,in_training_update_grad_gamma_beta_gen0407_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[127, 2, 7, 257], [127, 2, 7, 257]]",{},PASS(双输出cross_check),PASS(GPU参考),39.2,13.32,2.942943,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
31-geir,in_training_update_grad_gamma_beta_nchw_multiscale_original,float32,"NCHW (7, 8, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS37+geir,in_training_update_grad_gamma_beta_gen0427_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[5, 31, 7, 32], [5, 31, 7, 32]]",{},PASS(双输出cross_check),PASS(GPU参考),15.4,5.984,2.573529,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
32-geir,in_training_update_grad_gamma_beta_nhwc_multiscale_prefix,float32,"NHWC (5, 1, 1, 13)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS38+geir,in_training_update_grad_gamma_beta_gen0449_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 4, 32, 647], [3, 4, 32, 647]]",{},PASS(双输出cross_check),PASS(GPU参考),17.8,7.744,2.298554,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
33-geir,in_training_update_grad_gamma_beta_nhwc_multiscale_four_levels,float32,"NHWC (7, 1, 1, 17)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS39+geir,in_training_update_grad_gamma_beta_gen0469_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[63, 4, 15, 16], [63, 4, 15, 16]]",{},PASS(双输出cross_check),PASS(GPU参考),27.1,7.274,3.725598,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
34-geir,in_training_update_grad_gamma_beta_nd_multiscale_permutations,float32,"ND (7, 5040, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS40+geir,in_training_update_grad_gamma_beta_gen0489_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[5, 31, 5, 3], [5, 31, 5, 3]]",{},PASS(双输出cross_check),PASS(GPU参考),11.5,5.744,2.002089,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
35-geir,in_training_update_grad_gamma_beta_ncdhw_multiscale_full_range,float32,"NCDHW (25, 13, 1, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS41+geir,in_training_update_grad_gamma_beta_gen0510_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[513, 7, 2, 127], [513, 7, 2, 127]]",{},PASS(双输出cross_check),PASS(GPU参考),30.1,24.67,1.220105,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
36-geir,in_training_update_grad_gamma_beta_ndhwc_multiscale_mixed_columns,float32,"NDHWC (7, 1, 1, 1, 65)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS42+geir,in_training_update_grad_gamma_beta_gen0530_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[4, 3, 31, 257], [4, 3, 31, 257]]",{},PASS(双输出cross_check),PASS(GPU参考),14.6,6.11,2.389525,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
37-geir,in_training_update_grad_gamma_beta_nchw_multiscale_row_tiles,float32,"NCHW (8193, 17, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS43+geir,in_training_update_grad_gamma_beta_gen0550_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[127, 2, 1, 127], [127, 2, 1, 127]]",{},PASS(双输出cross_check),PASS(GPU参考),33.5,8.59,3.899884,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
38-geir,in_training_update_grad_gamma_beta_nchw_multiscale_column_tiles,float32,"NCHW (7, 400001, 1, 1)",{},PASS(NPU release binary双输出=PASS;CPU_Golden=PASS),不涉及（本档仅验证GEIR）,,,,,PASS44+geir,in_training_update_grad_gamma_beta_gen0571_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 4, 3, 15], [3, 4, 3, 15]]",{},PASS(双输出cross_check),PASS(GPU参考),11.4,4.608,2.473958,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
45+geir,in_training_update_grad_gamma_beta_gen0591_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[2, 7, 4, 65], [2, 7, 4, 65]]",{},PASS(双输出cross_check),PASS(GPU参考),14.5,5.332,2.71943,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
46+geir,in_training_update_grad_gamma_beta_gen0611_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[2, 2, 7, 16], [2, 2, 7, 16]]",{},PASS(双输出cross_check),PASS(GPU参考),11.2,4.632,2.417962,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
47+geir,in_training_update_grad_gamma_beta_gen0631_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[63, 7, 7, 1], [63, 7, 7, 1]]",{},PASS(双输出cross_check),PASS(GPU参考),20.6,6.72,3.065476,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
48+geir,in_training_update_grad_gamma_beta_gen0652_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[127, 8, 7, 8], [127, 8, 7, 8]]",{},PASS(双输出cross_check),PASS(GPU参考),21.8,9.03,2.414175,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
49+geir,in_training_update_grad_gamma_beta_gen0672_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 2, 7, 127], [1024, 2, 7, 127]]",{},PASS(双输出cross_check),PASS(GPU参考),45.4,43.752,1.037667,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
50+geir,in_training_update_grad_gamma_beta_gen0692_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[3, 4, 16, 4], [3, 4, 16, 4]]",{},PASS(双输出cross_check),PASS(GPU参考),11.4,4.378,2.603929,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
51+geir,in_training_update_grad_gamma_beta_gen0714_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 2, 4, 32], [1024, 2, 4, 32]]",{},PASS(双输出cross_check),PASS(GPU参考),33.5,36.576,0.915901,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
52+geir,in_training_update_grad_gamma_beta_gen0734_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[31, 8, 2, 64], [31, 8, 2, 64]]",{},PASS(双输出cross_check),PASS(GPU参考),20.7,6.744,3.069395,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
53+geir,in_training_update_grad_gamma_beta_gen0754_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 2, 15, 8], [1024, 2, 15, 8]]",{},PASS(双输出cross_check),PASS(GPU参考),33.3,36.484,0.912729,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
54+geir,in_training_update_grad_gamma_beta_gen0775_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 16, 31, 31], [16, 16, 31, 31]]",{},PASS(双输出cross_check),PASS(GPU参考),18.1,7.948,2.277302,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
55+geir,in_training_update_grad_gamma_beta_gen0795_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[8, 7, 7, 64], [8, 7, 7, 64]]",{},PASS(双输出cross_check),PASS(GPU参考),15.7,6.132,2.560339,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
56+geir,in_training_update_grad_gamma_beta_gen0815_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[4, 2, 7, 15], [4, 2, 7, 15]]",{},PASS(双输出cross_check),PASS(GPU参考),10.5,4.812,2.182045,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
57+geir,in_training_update_grad_gamma_beta_gen0836_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[8, 8, 15, 647], [8, 8, 15, 647]]",{},PASS(双输出cross_check),PASS(GPU参考),19.3,12.312,1.567576,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
58+geir,in_training_update_grad_gamma_beta_gen0856_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 4, 5, 5], [1024, 4, 5, 5]]",{},PASS(双输出cross_check),PASS(GPU参考),39.7,36.622,1.084048,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
59+geir,in_training_update_grad_gamma_beta_gen0876_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 7, 8, 1], [257, 7, 8, 1]]",{},PASS(双输出cross_check),PASS(GPU参考),18.6,12.128,1.533641,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
60+geir,in_training_update_grad_gamma_beta_gen0896_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 31, 1, 647], [257, 31, 1, 647]]",{},PASS(双输出cross_check),PASS(GPU参考),65.3,70.782,0.922551,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
61+geir,in_training_update_grad_gamma_beta_gen0917_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 4, 8, 649], [3, 4, 8, 649]]",{},PASS(双输出cross_check),PASS(GPU参考),14.9,6.348,2.347196,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
62+geir,in_training_update_grad_gamma_beta_gen0937_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[63, 5, 31, 257], [63, 5, 31, 257]]",{},PASS(双输出cross_check),PASS(GPU参考),32.2,36.73,0.876668,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
63+geir,in_training_update_grad_gamma_beta_gen0957_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[32, 16, 2, 31], [32, 16, 2, 31]]",{},PASS(双输出cross_check),PASS(GPU参考),20.8,6.308,3.2974,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
64+geir,in_training_update_grad_gamma_beta_gen0979_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1024, 2, 8, 7], [1024, 2, 8, 7]]",{},PASS(双输出cross_check),PASS(GPU参考),39.8,35.816,1.111235,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
65+geir,in_training_update_grad_gamma_beta_gen0999_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1, 5, 7, 7], [1, 5, 7, 7]]",{},PASS(双输出cross_check),PASS(GPU参考),13.1,4.672,2.803938,N/A,PASS(三方精度及SQLite审计；非TTK原生性能PASS)
Mchangwei_deliverables/september/INTrainingUpdateGradGammaBeta/05_test_result/performance/INTrainingUpdateGradGammaBeta_performance.csv+210-206
@@ -1,209 +1,213 @@
1项,内容1项,内容
2-算子,INTrainingUpdateGradGammaBeta（Ascend950）2+支持通路范围,仅支持Kernel直调与GEIR；ACLNN/Torch算子接口不支持；Torch-A100仅第三方参考。Kernel结果另见泛化/精度/性能专档。
3-口径,同一 200 条公共 4D shape；A5 为 TTK release binary bin_perf_us（预热后 10 次）；A100 为 eager/composite 的两次原生 torch.sum，使用 Nsight 统计每个 NVTX 范围内 CUDA kernel 总耗时（3 次预热、5 次测量）。3+证据版本,归档:evidence/perf_final_20260927/；kernel独立有效批次perf_20260927_retry1，该通路输入、结果、审计和DB分别核对；不同通路不混算，invalid批次不纳入。
4-结论,A5/A100 各 200/200 数值有效；A5 p50=4.682us，A100 p50=15.443us；A5 178 条更快，A100 22 条更快。4+判据,双输出stat_rel_err精度PASS；GPU torch device_us/NPU SQLite kernel us逐例>0.1。
5-门槛,逐例 GPU/A5 必须 >0.1；200/200 PASS，最小值=0.554458。除该底线外不设置额外性能合入门槛。5+覆盖,kernel 200个唯一同输入用例；结果不证明未覆盖设备/shape。
6-累加规则,复用已验证且与 200 条矩阵逐字段一致的 100 条，只补跑其精确补集 100 条；旧、新身份无重叠，按 200 条源矩阵顺序合并。6+结论,200/200精度PASS、数据库时延比值达标；最小G/N=0.540304。
7-审计,performance-audit-200.json：case_count=200、ratio_threshold_fail=0、result=PASS；A100 每条每次测量恰好 2 个 reduction kernel。GPU 端无同语义 fused/native 单接口，故本数据不宣称为融合算子上限。7+取数方法,GPU取torch profiler device_us；NPU取ai_core_op_summary.db/task_time纳秒转微秒，Kernel丢首条后取中位数，GEIR全部取中位数。
8+TTK原始状态,UNKNOWN 200/200；不是TTK原生性能PASS。
9+折算比,folded_G/N为N/A：没有可靠硬件折算因子；不以raw G/N冒充折算值。
10+运行模式,--pc=1；Kernel --binary-only等价--dynamic=false --binary=release；GEIR为const图执行。
11+环境限制,仅Ascend950PR_9579实测；私有CANN二进制补丁不等同正式源码构建验证；不构成完整CI放行。
8 12 
9testcase,dtype,shape,attributes,tilingKey,npu_perf_us,gpu_us,raw_G/N,folded_G/N13testcase,dtype,shape,attributes,tilingKey,npu_perf_us,gpu_us,raw_G/N,folded_G/N
10-in_training_update_grad_gamma_beta_gen0000_nchw_f32,float32,"NCHW [1, 1, 1, 1]",{},0,1.670000,8.857400,5.303832,5.30383214+in_training_update_grad_gamma_beta_gen0000_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1, 1, 1, 1], [1, 1, 1, 1]]",{},0 (0x0000000000000000),1.833,10.5,5.728314,N/A
11-in_training_update_grad_gamma_beta_gen0005_nhwc_f32,float32,"NHWC [3, 1, 1, 7]",{},0,2.116000,9.126200,4.312949,4.31294915+in_training_update_grad_gamma_beta_gen0005_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 1, 1, 7], [3, 1, 1, 7]]",{},0 (0x0000000000000000),2.238,10.8,4.825737,N/A
12-in_training_update_grad_gamma_beta_gen0010_nchw_f32,float32,"NCHW [31, 1, 1, 63]",{},0,4.232000,13.279600,3.137902,3.13790216+in_training_update_grad_gamma_beta_gen0010_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[31, 1, 1, 63], [31, 1, 1, 63]]",{},0 (0x0000000000000000),4.405,15.1,3.427923,N/A
13-in_training_update_grad_gamma_beta_gen0015_nhwc_f32,float32,"NHWC [65, 1, 1, 65]",{},0,5.186000,18.649600,3.596143,3.59614317+in_training_update_grad_gamma_beta_gen0015_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[65, 1, 1, 65], [65, 1, 1, 65]]",{},0 (0x0000000000000000),5.312,21.1,3.972139,N/A
14-in_training_update_grad_gamma_beta_gen0020_nchw_f32,float32,"NCHW [2, 1, 1, 648]",{},0,3.024000,12.454000,4.118386,4.11838618+in_training_update_grad_gamma_beta_gen0020_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 1, 1, 648], [2, 1, 1, 648]]",{},0 (0x0000000000000000),3.086,14.2,4.601426,N/A
15-in_training_update_grad_gamma_beta_gen0025_nhwc_f32,float32,"NHWC [3, 2, 5, 7]",{},0,3.116000,9.638200,3.093132,3.09313219+in_training_update_grad_gamma_beta_gen0025_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 2, 5, 7], [3, 2, 5, 7]]",{},0 (0x0000000000000000),3.078,11.0,3.573749,N/A
16-in_training_update_grad_gamma_beta_gen0030_nchw_f32,float32,"NCHW [1024, 3, 5, 7]",{},0,36.544000,24.447400,0.668985,0.66898520+in_training_update_grad_gamma_beta_gen0030_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 3, 5, 7], [1024, 3, 5, 7]]",{},0 (0x0000000000000000),36.46,27.6,0.756994,N/A
17-in_training_update_grad_gamma_beta_gen0035_nhwc_f32,float32,"NHWC [1024, 2, 8, 2]",{},0,35.050000,22.431600,0.639989,0.63998921+in_training_update_grad_gamma_beta_gen0035_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1024, 2, 8, 2], [1024, 2, 8, 2]]",{},0 (0x0000000000000000),35.347,23.6,0.667666,N/A
18-in_training_update_grad_gamma_beta_gen0040_nchw_f32,float32,"NCHW [8, 7, 1, 64]",{},0,3.626000,10.976000,3.027027,3.02702722+in_training_update_grad_gamma_beta_gen0040_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[8, 7, 1, 64], [8, 7, 1, 64]]",{},0 (0x0000000000000000),3.645,12.5,3.429355,N/A
19-in_training_update_grad_gamma_beta_gen0045_nhwc_f32,float32,"NHWC [257, 1, 8, 65]",{},0,11.068000,24.403200,2.204843,2.20484323+in_training_update_grad_gamma_beta_gen0045_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[257, 1, 8, 65], [257, 1, 8, 65]]",{},0 (0x0000000000000000),11.368,27.8,2.445461,N/A
20-in_training_update_grad_gamma_beta_gen0050_nchw_f32,float32,"NCHW [1, 31, 8, 1]",{},0,2.402000,12.563200,5.230308,5.23030824+in_training_update_grad_gamma_beta_gen0050_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1, 31, 8, 1], [1, 31, 8, 1]]",{},0 (0x0000000000000000),2.452,14.1,5.750408,N/A
21-in_training_update_grad_gamma_beta_gen0055_nhwc_f32,float32,"NHWC [16, 2, 5, 63]",{},0,3.802000,13.830200,3.637612,3.63761225+in_training_update_grad_gamma_beta_gen0055_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 2, 5, 63], [16, 2, 5, 63]]",{},0 (0x0000000000000000),3.911,15.7,4.014319,N/A
22-in_training_update_grad_gamma_beta_gen0060_nchw_f32,float32,"NCHW [31, 3, 7, 2]",{},0,4.100000,13.433600,3.276488,3.27648826+in_training_update_grad_gamma_beta_gen0060_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[31, 3, 7, 2], [31, 3, 7, 2]]",{},0 (0x0000000000000000),4.402,15.6,3.543844,N/A
23-in_training_update_grad_gamma_beta_gen0065_nhwc_f32,float32,"NHWC [63, 5, 1, 63]",{},0,5.148000,19.692800,3.825330,3.82533027+in_training_update_grad_gamma_beta_gen0065_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[63, 5, 1, 63], [63, 5, 1, 63]]",{},0 (0x0000000000000000),5.192,22.5,4.33359,N/A
24-in_training_update_grad_gamma_beta_gen0070_nchw_f32,float32,"NCHW [2, 16, 7, 8]",{},0,2.974000,12.543600,4.217754,4.21775428+in_training_update_grad_gamma_beta_gen0070_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 16, 7, 8], [2, 16, 7, 8]]",{},0 (0x0000000000000000),3.325,14.3,4.300752,N/A
25-in_training_update_grad_gamma_beta_gen0075_nhwc_f32,float32,"NHWC [1024, 16, 8, 31]",{},0,73.216000,40.595200,0.554458,0.55445829+in_training_update_grad_gamma_beta_gen0075_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1024, 16, 8, 31], [1024, 16, 8, 31]]",{},0 (0x0000000000000000),73.63,53.6,0.727964,N/A
26-in_training_update_grad_gamma_beta_gen0080_nchw_f32,float32,"NCHW [64, 8, 8, 7]",{},0,5.150000,15.948600,3.096816,3.09681630+in_training_update_grad_gamma_beta_gen0080_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[64, 8, 8, 7], [64, 8, 8, 7]]",{},0 (0x0000000000000000),5.415,18.2,3.361034,N/A
27-in_training_update_grad_gamma_beta_gen0085_nhwc_f32,float32,"NHWC [1, 31, 32, 1]",{},0,3.030000,12.480000,4.118812,4.11881231+in_training_update_grad_gamma_beta_gen0085_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1, 31, 32, 1], [1, 31, 32, 1]]",{},0 (0x0000000000000000),2.954,14.1,4.773189,N/A
28-in_training_update_grad_gamma_beta_gen0090_nchw_f32,float32,"NCHW [3, 2, 16, 5]",{},0,3.148000,10.016000,3.181703,3.18170332+in_training_update_grad_gamma_beta_gen0090_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[3, 2, 16, 5], [3, 2, 16, 5]]",{},0 (0x0000000000000000),3.124,11.4,3.649168,N/A
29-in_training_update_grad_gamma_beta_gen0095_nhwc_f32,float32,"NHWC [1, 4, 7, 647]",{},0,3.352000,12.748800,3.803341,3.80334133+in_training_update_grad_gamma_beta_gen0095_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1, 4, 7, 647], [1, 4, 7, 647]]",{},0 (0x0000000000000000),3.477,14.5,4.170262,N/A
30-in_training_update_grad_gamma_beta_gen0100_nchw_f32,float32,"NCHW [513, 2, 4, 649]",{},0,42.176000,38.700400,0.917593,0.91759334+in_training_update_grad_gamma_beta_gen0100_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[513, 2, 4, 649], [513, 2, 4, 649]]",{},0 (0x0000000000000000),42.338,50.3,1.188058,N/A
31-in_training_update_grad_gamma_beta_gen0105_nhwc_f32,float32,"NHWC [15, 7, 32, 32]",{},0,3.986000,15.590200,3.911239,3.91123935+in_training_update_grad_gamma_beta_gen0105_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[15, 7, 32, 32], [15, 7, 32, 32]]",{},0 (0x0000000000000000),4.18,17.8,4.258373,N/A
32-in_training_update_grad_gamma_beta_gen0110_nchw_f32,float32,"NCHW [5, 7, 31, 15]",{},0,4.062000,12.921200,3.180995,3.18099536+in_training_update_grad_gamma_beta_gen0110_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[5, 7, 31, 15], [5, 7, 31, 15]]",{},0 (0x0000000000000000),4.189,14.8,3.533063,N/A
33-in_training_update_grad_gamma_beta_gen0115_nhwc_f32,float32,"NHWC [7, 3, 32, 7]",{},0,3.384000,13.394600,3.958215,3.95821537+in_training_update_grad_gamma_beta_gen0115_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[7, 3, 32, 7], [7, 3, 32, 7]]",{},0 (0x0000000000000000),3.595,15.2,4.228095,N/A
34-in_training_update_grad_gamma_beta_gen0120_nchw_f32,float32,"NCHW [15, 5, 2, 648]",{},0,4.700000,15.552000,3.308936,3.30893638+in_training_update_grad_gamma_beta_gen0120_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[15, 5, 2, 648], [15, 5, 2, 648]]",{},0 (0x0000000000000000),4.659,17.7,3.799099,N/A
35-in_training_update_grad_gamma_beta_gen0126_nchw_f32,float32,"NCHW [127, 1, 15, 648]",{},0,18.488000,21.126000,1.142687,1.14268739+in_training_update_grad_gamma_beta_gen0126_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[127, 1, 15, 648], [127, 1, 15, 648]]",{},0 (0x0000000000000000),18.788,24.1,1.282734,N/A
36-in_training_update_grad_gamma_beta_gen0131_nhwc_f32,float32,"NHWC [32, 16, 4, 127]",{},0,6.708000,18.739200,2.793560,2.79356040+in_training_update_grad_gamma_beta_gen0131_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[32, 16, 4, 127], [32, 16, 4, 127]]",{},0 (0x0000000000000000),7.047,21.4,3.036753,N/A
37-in_training_update_grad_gamma_beta_gen0136_nchw_f32,float32,"NCHW [15, 4, 1, 16]",{},0,3.554000,11.904000,3.349465,3.34946541+in_training_update_grad_gamma_beta_gen0136_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[15, 4, 1, 16], [15, 4, 1, 16]]",{},0 (0x0000000000000000),3.676,13.7,3.726877,N/A
38-in_training_update_grad_gamma_beta_gen0141_nhwc_f32,float32,"NHWC [4, 8, 16, 127]",{},0,3.992000,13.190400,3.304208,3.30420842+in_training_update_grad_gamma_beta_gen0141_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[4, 8, 16, 127], [4, 8, 16, 127]]",{},0 (0x0000000000000000),4.224,15.0,3.551136,N/A
39-in_training_update_grad_gamma_beta_gen0146_nchw_f32,float32,"NCHW [32, 3, 4, 65]",{},0,4.326000,17.811000,4.117198,4.11719843+in_training_update_grad_gamma_beta_gen0146_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[32, 3, 4, 65], [32, 3, 4, 65]]",{},0 (0x0000000000000000),4.549,20.1,4.418554,N/A
40-in_training_update_grad_gamma_beta_gen0151_nhwc_f32,float32,"NHWC [257, 16, 32, 63]",{},0,93.434000,63.314400,0.677638,0.67763844+in_training_update_grad_gamma_beta_gen0151_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[257, 16, 32, 63], [257, 16, 32, 63]]",{},0 (0x0000000000000000),94.392,66.9,0.708747,N/A
41-in_training_update_grad_gamma_beta_gen0156_nchw_f32,float32,"NCHW [3, 8, 3, 63]",{},0,3.380000,12.806400,3.788876,3.78887645+in_training_update_grad_gamma_beta_gen0156_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[3, 8, 3, 63], [3, 8, 3, 63]]",{},0 (0x0000000000000000),3.43,14.6,4.25656,N/A
42-in_training_update_grad_gamma_beta_gen0161_nhwc_f32,float32,"NHWC [3, 7, 3, 7]",{},0,3.108000,8.774400,2.823166,2.82316646+in_training_update_grad_gamma_beta_gen0161_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 7, 3, 7], [3, 7, 3, 7]]",{},0 (0x0000000000000000),3.114,10.4,3.339756,N/A
43-in_training_update_grad_gamma_beta_gen0166_nchw_f32,float32,"NCHW [257, 5, 7, 1]",{},0,11.014000,15.974400,1.450372,1.45037247+in_training_update_grad_gamma_beta_gen0166_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 5, 7, 1], [257, 5, 7, 1]]",{},0 (0x0000000000000000),11.091,18.1,1.631954,N/A
44-in_training_update_grad_gamma_beta_gen0171_nhwc_f32,float32,"NHWC [5, 16, 16, 64]",{},0,4.300000,13.798400,3.208930,3.20893048+in_training_update_grad_gamma_beta_gen0171_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[5, 16, 16, 64], [5, 16, 16, 64]]",{},0 (0x0000000000000000),4.648,15.7,3.377797,N/A
45-in_training_update_grad_gamma_beta_gen0176_nchw_f32,float32,"NCHW [1, 8, 7, 649]",{},0,3.350000,12.985400,3.876239,3.87623949+in_training_update_grad_gamma_beta_gen0176_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1, 8, 7, 649], [1, 8, 7, 649]]",{},0 (0x0000000000000000),3.488,14.8,4.243119,N/A
46-in_training_update_grad_gamma_beta_gen0181_nhwc_f32,float32,"NHWC [16, 1, 5, 32]",{},0,3.666000,11.897400,3.245336,3.24533650+in_training_update_grad_gamma_beta_gen0181_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 1, 5, 32], [16, 1, 5, 32]]",{},0 (0x0000000000000000),3.732,13.6,3.644159,N/A
47-in_training_update_grad_gamma_beta_gen0186_nchw_f32,float32,"NCHW [257, 1, 2, 649]",{},0,12.536000,19.654400,1.567837,1.56783751+in_training_update_grad_gamma_beta_gen0186_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 1, 2, 649], [257, 1, 2, 649]]",{},0 (0x0000000000000000),12.63,22.4,1.773555,N/A
48-in_training_update_grad_gamma_beta_gen0191_nhwc_f32,float32,"NHWC [4, 31, 31, 32]",{},0,4.792000,13.375800,2.791277,2.79127752+in_training_update_grad_gamma_beta_gen0191_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[4, 31, 31, 32], [4, 31, 31, 32]]",{},0 (0x0000000000000000),4.799,15.2,3.167327,N/A
49-in_training_update_grad_gamma_beta_gen0196_nchw_f32,float32,"NCHW [5, 2, 32, 649]",{},0,5.402000,14.431600,2.671529,2.67152953+in_training_update_grad_gamma_beta_gen0196_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[5, 2, 32, 649], [5, 2, 32, 649]]",{},0 (0x0000000000000000),5.471,16.4,2.997624,N/A
50-in_training_update_grad_gamma_beta_gen0201_nhwc_f32,float32,"NHWC [2, 1, 15, 15]",{},0,2.696000,9.203200,3.413650,3.41365054+in_training_update_grad_gamma_beta_gen0201_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[2, 1, 15, 15], [2, 1, 15, 15]]",{},0 (0x0000000000000000),2.765,10.5,3.797468,N/A
51-in_training_update_grad_gamma_beta_gen0206_nchw_f32,float32,"NCHW [32, 16, 15, 649]",{},0,53.172000,44.607400,0.838927,0.83892755+in_training_update_grad_gamma_beta_gen0206_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[32, 16, 15, 649], [32, 16, 15, 649]]",{},0 (0x0000000000000000),52.452,48.1,0.917029,N/A
52-in_training_update_grad_gamma_beta_gen0211_nhwc_f32,float32,"NHWC [63, 3, 15, 1]",{},0,5.090000,18.111800,3.558310,3.55831056+in_training_update_grad_gamma_beta_gen0211_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[63, 3, 15, 1], [63, 3, 15, 1]]",{},0 (0x0000000000000000),5.215,20.6,3.950144,N/A
53-in_training_update_grad_gamma_beta_gen0216_nchw_f32,float32,"NCHW [5, 8, 5, 32]",{},0,3.482000,13.343600,3.832165,3.83216557+in_training_update_grad_gamma_beta_gen0216_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[5, 8, 5, 32], [5, 8, 5, 32]]",{},0 (0x0000000000000000),3.465,15.2,4.386724,N/A
54-in_training_update_grad_gamma_beta_gen0221_nhwc_f32,float32,"NHWC [32, 3, 16, 63]",{},0,4.930000,18.278400,3.707586,3.70758658+in_training_update_grad_gamma_beta_gen0221_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[32, 3, 16, 63], [32, 3, 16, 63]]",{},0 (0x0000000000000000),5.0,20.9,4.18,N/A
55-in_training_update_grad_gamma_beta_gen0226_nchw_f32,float32,"NCHW [64, 4, 5, 3]",{},0,5.104000,19.730800,3.865752,3.86575259+in_training_update_grad_gamma_beta_gen0226_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[64, 4, 5, 3], [64, 4, 5, 3]]",{},0 (0x0000000000000000),5.298,22.4,4.228011,N/A
56-in_training_update_grad_gamma_beta_gen0231_nhwc_f32,float32,"NHWC [65, 8, 3, 1]",{},0,5.040000,20.121600,3.992381,3.99238160+in_training_update_grad_gamma_beta_gen0231_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[65, 8, 3, 1], [65, 8, 3, 1]]",{},0 (0x0000000000000000),5.073,22.8,4.494382,N/A
57-in_training_update_grad_gamma_beta_gen0236_nchw_f32,float32,"NCHW [63, 1, 5, 32]",{},0,5.056000,20.230400,4.001266,4.00126661+in_training_update_grad_gamma_beta_gen0236_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[63, 1, 5, 32], [63, 1, 5, 32]]",{},0 (0x0000000000000000),5.226,22.9,4.381936,N/A
58-in_training_update_grad_gamma_beta_gen0241_nhwc_f32,float32,"NHWC [257, 31, 2, 8]",{},0,11.162000,24.646200,2.208045,2.20804562+in_training_update_grad_gamma_beta_gen0241_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[257, 31, 2, 8], [257, 31, 2, 8]]",{},0 (0x0000000000000000),11.24,28.2,2.508897,N/A
59-in_training_update_grad_gamma_beta_gen0246_nchw_f32,float32,"NCHW [32, 4, 5, 7]",{},0,4.324000,14.553400,3.365726,3.36572663+in_training_update_grad_gamma_beta_gen0246_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[32, 4, 5, 7], [32, 4, 5, 7]]",{},0 (0x0000000000000000),4.356,16.5,3.787879,N/A
60-in_training_update_grad_gamma_beta_gen0251_nhwc_f32,float32,"NHWC [2, 4, 15, 1]",{},0,2.676000,10.079800,3.766741,3.76674164+in_training_update_grad_gamma_beta_gen0251_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[2, 4, 15, 1], [2, 4, 15, 1]]",{},0 (0x0000000000000000),2.706,11.6,4.28677,N/A
61-in_training_update_grad_gamma_beta_gen0256_nchw_f32,float32,"NCHW [257, 1, 31, 8]",{},0,10.860000,24.358200,2.242928,2.24292865+in_training_update_grad_gamma_beta_gen0256_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 1, 31, 8], [257, 1, 31, 8]]",{},0 (0x0000000000000000),10.923,27.6,2.526778,N/A
62-in_training_update_grad_gamma_beta_gen0261_nhwc_f32,float32,"NHWC [64, 4, 16, 5]",{},0,4.976000,15.814200,3.178095,3.17809566+in_training_update_grad_gamma_beta_gen0261_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[64, 4, 16, 5], [64, 4, 16, 5]]",{},0 (0x0000000000000000),5.153,18.1,3.512517,N/A
63-in_training_update_grad_gamma_beta_gen0266_nchw_f32,float32,"NCHW [16, 8, 31, 31]",{},0,5.180000,15.596600,3.010927,3.01092767+in_training_update_grad_gamma_beta_gen0266_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[16, 8, 31, 31], [16, 8, 31, 31]]",{},0 (0x0000000000000000),5.435,17.7,3.25667,N/A
64-in_training_update_grad_gamma_beta_gen0271_nhwc_f32,float32,"NHWC [2, 4, 4, 649]",{},0,3.336000,12.736000,3.817746,3.81774668+in_training_update_grad_gamma_beta_gen0271_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[2, 4, 4, 649], [2, 4, 4, 649]]",{},0 (0x0000000000000000),3.501,14.6,4.170237,N/A
65-in_training_update_grad_gamma_beta_gen0276_nchw_f32,float32,"NCHW [65, 5, 15, 3]",{},0,5.144000,19.967800,3.881765,3.88176569+in_training_update_grad_gamma_beta_gen0276_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[65, 5, 15, 3], [65, 5, 15, 3]]",{},0 (0x0000000000000000),5.404,22.4,4.145078,N/A
66-in_training_update_grad_gamma_beta_gen0281_nhwc_f32,float32,"NHWC [7, 2, 5, 64]",{},0,3.496000,13.344000,3.816934,3.81693470+in_training_update_grad_gamma_beta_gen0281_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[7, 2, 5, 64], [7, 2, 5, 64]]",{},0 (0x0000000000000000),3.617,15.2,4.202378,N/A
67-in_training_update_grad_gamma_beta_gen0286_nchw_f32,float32,"NCHW [127, 3, 32, 8]",{},0,6.934000,19.200000,2.768965,2.76896571+in_training_update_grad_gamma_beta_gen0286_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[127, 3, 32, 8], [127, 3, 32, 8]]",{},0 (0x0000000000000000),6.959,21.7,3.118264,N/A
68-in_training_update_grad_gamma_beta_gen0291_nhwc_f32,float32,"NHWC [127, 2, 15, 257]",{},0,17.382000,35.602800,2.048257,2.04825772+in_training_update_grad_gamma_beta_gen0291_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[127, 2, 15, 257], [127, 2, 15, 257]]",{},0 (0x0000000000000000),17.577,40.7,2.315526,N/A
69-in_training_update_grad_gamma_beta_gen0296_nchw_f32,float32,"NCHW [15, 7, 2, 649]",{},0,5.086000,15.481200,3.043885,3.04388573+in_training_update_grad_gamma_beta_gen0296_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[15, 7, 2, 649], [15, 7, 2, 649]]",{},0 (0x0000000000000000),5.094,17.7,3.474676,N/A
70-in_training_update_grad_gamma_beta_gen0301_nhwc_f32,float32,"NHWC [16, 1, 32, 64]",{},0,3.974000,15.238200,3.834474,3.83447474+in_training_update_grad_gamma_beta_gen0301_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 1, 32, 64], [16, 1, 32, 64]]",{},0 (0x0000000000000000),4.082,17.4,4.262616,N/A
71-in_training_update_grad_gamma_beta_gen0306_nchw_f32,float32,"NCHW [4, 1, 3, 1]",{},0,2.264000,9.004800,3.977385,3.97738575+in_training_update_grad_gamma_beta_gen0306_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[4, 1, 3, 1], [4, 1, 3, 1]]",{},0 (0x0000000000000000),2.252,10.6,4.706927,N/A
72-in_training_update_grad_gamma_beta_gen0311_nhwc_f32,float32,"NHWC [15, 31, 32, 5]",{},0,4.810000,15.500600,3.222578,3.22257876+in_training_update_grad_gamma_beta_gen0311_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[15, 31, 32, 5], [15, 31, 32, 5]]",{},0 (0x0000000000000000),4.94,17.7,3.582996,N/A
73-in_training_update_grad_gamma_beta_gen0316_nchw_f32,float32,"NCHW [1024, 8, 15, 3]",{},0,34.776000,26.777200,0.769991,0.76999177+in_training_update_grad_gamma_beta_gen0316_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 8, 15, 3], [1024, 8, 15, 3]]",{},0 (0x0000000000000000),35.272,34.8,0.986618,N/A
74-in_training_update_grad_gamma_beta_gen0321_nhwc_f32,float32,"NHWC [15, 4, 7, 127]",{},0,4.284000,15.449600,3.606349,3.60634978+in_training_update_grad_gamma_beta_gen0321_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[15, 4, 7, 127], [15, 4, 7, 127]]",{},0 (0x0000000000000000),4.415,17.6,3.98641,N/A
75-in_training_update_grad_gamma_beta_gen0326_nchw_f32,float32,"NCHW [3, 31, 5, 65]",{},0,4.002000,12.531000,3.131184,3.13118479+in_training_update_grad_gamma_beta_gen0326_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[3, 31, 5, 65], [3, 31, 5, 65]]",{},0 (0x0000000000000000),4.137,14.4,3.480783,N/A
76-in_training_update_grad_gamma_beta_gen0331_nhwc_f32,float32,"NHWC [8, 4, 1, 16]",{},0,3.356000,10.195200,3.037902,3.03790280+in_training_update_grad_gamma_beta_gen0331_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[8, 4, 1, 16], [8, 4, 1, 16]]",{},0 (0x0000000000000000),3.483,12.0,3.445306,N/A
77-in_training_update_grad_gamma_beta_gen0336_nchw_f32,float32,"NCHW [4, 7, 31, 4]",{},0,3.486000,12.716800,3.647963,3.64796381+in_training_update_grad_gamma_beta_gen0336_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[4, 7, 31, 4], [4, 7, 31, 4]]",{},0 (0x0000000000000000),3.548,14.4,4.058625,N/A
78-in_training_update_grad_gamma_beta_gen0341_nhwc_f32,float32,"NHWC [8, 5, 8, 3]",{},0,3.322000,10.284600,3.095906,3.09590682+in_training_update_grad_gamma_beta_gen0341_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[8, 5, 8, 3], [8, 5, 8, 3]]",{},0 (0x0000000000000000),3.409,12.1,3.549428,N/A
79-in_training_update_grad_gamma_beta_gen0346_nchw_f32,float32,"NCHW [31, 2, 32, 31]",{},0,4.280000,18.374400,4.293084,4.29308483+in_training_update_grad_gamma_beta_gen0346_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[31, 2, 32, 31], [31, 2, 32, 31]]",{},0 (0x0000000000000000),4.538,20.9,4.605553,N/A
80-in_training_update_grad_gamma_beta_gen0351_nhwc_f32,float32,"NHWC [1, 31, 5, 64]",{},0,3.226000,12.658800,3.923993,3.92399384+in_training_update_grad_gamma_beta_gen0351_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1, 31, 5, 64], [1, 31, 5, 64]]",{},0 (0x0000000000000000),3.125,14.4,4.608,N/A
81-in_training_update_grad_gamma_beta_gen0356_nchw_f32,float32,"NCHW [2, 4, 7, 1]",{},0,2.416000,10.105600,4.182781,4.18278185+in_training_update_grad_gamma_beta_gen0356_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 4, 7, 1], [2, 4, 7, 1]]",{},0 (0x0000000000000000),2.351,11.5,4.891536,N/A
82-in_training_update_grad_gamma_beta_gen0361_nhwc_f32,float32,"NHWC [513, 4, 8, 2]",{},0,18.540000,22.816000,1.230636,1.23063686+in_training_update_grad_gamma_beta_gen0361_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[513, 4, 8, 2], [513, 4, 8, 2]]",{},0 (0x0000000000000000),19.21,25.5,1.327434,N/A
83-in_training_update_grad_gamma_beta_gen0366_nchw_f32,float32,"NCHW [16, 7, 5, 63]",{},0,4.270000,14.668600,3.435269,3.43526987+in_training_update_grad_gamma_beta_gen0366_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[16, 7, 5, 63], [16, 7, 5, 63]]",{},0 (0x0000000000000000),4.416,16.6,3.759058,N/A
84-in_training_update_grad_gamma_beta_gen0371_nhwc_f32,float32,"NHWC [65, 1, 32, 15]",{},0,5.244000,16.236800,3.096262,3.09626288+in_training_update_grad_gamma_beta_gen0371_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[65, 1, 32, 15], [65, 1, 32, 15]]",{},0 (0x0000000000000000),5.327,18.4,3.454102,N/A
85-in_training_update_grad_gamma_beta_gen0377_nhwc_f32,float32,"NHWC [31, 1, 8, 31]",{},0,3.940000,15.091000,3.830203,3.83020389+in_training_update_grad_gamma_beta_gen0377_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[31, 1, 8, 31], [31, 1, 8, 31]]",{},0 (0x0000000000000000),4.055,17.1,4.217016,N/A
86-in_training_update_grad_gamma_beta_gen0382_nchw_f32,float32,"NCHW [513, 1, 7, 1]",{},0,17.582000,14.739000,0.838301,0.83830190+in_training_update_grad_gamma_beta_gen0382_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[513, 1, 7, 1], [513, 1, 7, 1]]",{},0 (0x0000000000000000),18.185,16.8,0.923838,N/A
87-in_training_update_grad_gamma_beta_gen0387_nhwc_f32,float32,"NHWC [257, 16, 16, 15]",{},0,20.184000,26.323000,1.304152,1.30415291+in_training_update_grad_gamma_beta_gen0387_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[257, 16, 16, 15], [257, 16, 16, 15]]",{},0 (0x0000000000000000),20.244,30.1,1.48686,N/A
88-in_training_update_grad_gamma_beta_gen0392_nchw_f32,float32,"NCHW [64, 31, 15, 127]",{},0,46.768000,42.098400,0.900154,0.90015492+in_training_update_grad_gamma_beta_gen0392_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[64, 31, 15, 127], [64, 31, 15, 127]]",{},0 (0x0000000000000000),46.92,48.5,1.033674,N/A
89-in_training_update_grad_gamma_beta_gen0397_nhwc_f32,float32,"NHWC [65, 5, 1, 15]",{},0,5.242000,18.649400,3.557688,3.55768893+in_training_update_grad_gamma_beta_gen0397_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[65, 5, 1, 15], [65, 5, 1, 15]]",{},0 (0x0000000000000000),5.373,21.2,3.945654,N/A
90-in_training_update_grad_gamma_beta_gen0402_nchw_f32,float32,"NCHW [4, 8, 5, 31]",{},0,3.560000,12.876800,3.617079,3.61707994+in_training_update_grad_gamma_beta_gen0402_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[4, 8, 5, 31], [4, 8, 5, 31]]",{},0 (0x0000000000000000),3.497,13.6,3.889048,N/A
91-in_training_update_grad_gamma_beta_gen0407_nhwc_f32,float32,"NHWC [127, 2, 7, 257]",{},0,11.112000,34.329200,3.089381,3.08938195+in_training_update_grad_gamma_beta_gen0407_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[127, 2, 7, 257], [127, 2, 7, 257]]",{},0 (0x0000000000000000),11.187,39.1,3.495128,N/A
92-in_training_update_grad_gamma_beta_gen0412_nchw_f32,float32,"NCHW [127, 4, 1, 127]",{},0,7.110000,18.931200,2.662616,2.66261696+in_training_update_grad_gamma_beta_gen0412_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[127, 4, 1, 127], [127, 4, 1, 127]]",{},0 (0x0000000000000000),7.33,21.6,2.946794,N/A
93-in_training_update_grad_gamma_beta_gen0417_nhwc_f32,float32,"NHWC [1024, 4, 31, 7]",{},0,38.108000,28.632800,0.751359,0.75135997+in_training_update_grad_gamma_beta_gen0417_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1024, 4, 31, 7], [1024, 4, 31, 7]]",{},0 (0x0000000000000000),38.556,35.9,0.931113,N/A
94-in_training_update_grad_gamma_beta_gen0422_nchw_f32,float32,"NCHW [513, 8, 8, 5]",{},0,18.814000,36.140400,1.920931,1.92093198+in_training_update_grad_gamma_beta_gen0422_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[513, 8, 8, 5], [513, 8, 8, 5]]",{},0 (0x0000000000000000),18.989,41.1,2.164411,N/A
95-in_training_update_grad_gamma_beta_gen0427_nhwc_f32,float32,"NHWC [5, 31, 7, 32]",{},0,3.924000,13.484800,3.436493,3.43649399+in_training_update_grad_gamma_beta_gen0427_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[5, 31, 7, 32], [5, 31, 7, 32]]",{},0 (0x0000000000000000),4.01,15.4,3.840399,N/A
96-in_training_update_grad_gamma_beta_gen0432_nchw_f32,float32,"NCHW [2, 8, 4, 8]",{},0,2.688000,9.766000,3.633185,3.633185100+in_training_update_grad_gamma_beta_gen0432_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 8, 4, 8], [2, 8, 4, 8]]",{},0 (0x0000000000000000),2.868,11.1,3.870293,N/A
97-in_training_update_grad_gamma_beta_gen0437_nhwc_f32,float32,"NHWC [16, 3, 5, 2]",{},0,3.662000,10.931200,2.985035,2.985035101+in_training_update_grad_gamma_beta_gen0437_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 3, 5, 2], [16, 3, 5, 2]]",{},0 (0x0000000000000000),3.678,12.7,3.452964,N/A
98-in_training_update_grad_gamma_beta_gen0442_nchw_f32,float32,"NCHW [63, 31, 15, 7]",{},0,6.078000,23.251000,3.825436,3.825436102+in_training_update_grad_gamma_beta_gen0442_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[63, 31, 15, 7], [63, 31, 15, 7]]",{},0 (0x0000000000000000),6.182,26.4,4.270463,N/A
99-in_training_update_grad_gamma_beta_gen0447_nhwc_f32,float32,"NHWC [5, 8, 2, 7]",{},0,3.212000,10.080000,3.138232,3.138232103+in_training_update_grad_gamma_beta_gen0447_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[5, 8, 2, 7], [5, 8, 2, 7]]",{},0 (0x0000000000000000),3.303,11.9,3.602785,N/A
100-in_training_update_grad_gamma_beta_gen0452_nchw_f32,float32,"NCHW [513, 8, 8, 16]",{},0,19.780000,36.332400,1.836825,1.836825104+in_training_update_grad_gamma_beta_gen0452_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[513, 8, 8, 16], [513, 8, 8, 16]]",{},0 (0x0000000000000000),20.199,41.4,2.049606,N/A
101-in_training_update_grad_gamma_beta_gen0457_nhwc_f32,float32,"NHWC [513, 1, 32, 1]",{},0,19.162000,17.043000,0.889417,0.889417105+in_training_update_grad_gamma_beta_gen0457_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[513, 1, 32, 1], [513, 1, 32, 1]]",{},0 (0x0000000000000000),19.478,19.4,0.995995,N/A
102-in_training_update_grad_gamma_beta_gen0462_nchw_f32,float32,"NCHW [5, 2, 1, 648]",{},0,3.620000,13.382400,3.696796,3.696796106+in_training_update_grad_gamma_beta_gen0462_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[5, 2, 1, 648], [5, 2, 1, 648]]",{},0 (0x0000000000000000),3.568,15.2,4.26009,N/A
103-in_training_update_grad_gamma_beta_gen0467_nhwc_f32,float32,"NHWC [513, 3, 5, 8]",{},0,18.532000,24.377200,1.315411,1.315411107+in_training_update_grad_gamma_beta_gen0467_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[513, 3, 5, 8], [513, 3, 5, 8]]",{},0 (0x0000000000000000),19.133,27.9,1.458214,N/A
104-in_training_update_grad_gamma_beta_gen0472_nchw_f32,float32,"NCHW [63, 31, 32, 1]",{},0,5.164000,23.852400,4.618978,4.618978108+in_training_update_grad_gamma_beta_gen0472_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[63, 31, 32, 1], [63, 31, 32, 1]]",{},0 (0x0000000000000000),5.14,27.1,5.272374,N/A
105-in_training_update_grad_gamma_beta_gen0477_nhwc_f32,float32,"NHWC [4, 1, 31, 8]",{},0,3.156000,9.971200,3.159442,3.159442109+in_training_update_grad_gamma_beta_gen0477_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[4, 1, 31, 8], [4, 1, 31, 8]]",{},0 (0x0000000000000000),3.174,11.4,3.591682,N/A
106-in_training_update_grad_gamma_beta_gen0482_nchw_f32,float32,"NCHW [63, 16, 2, 3]",{},0,4.956000,20.108400,4.057385,4.057385110+in_training_update_grad_gamma_beta_gen0482_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[63, 16, 2, 3], [63, 16, 2, 3]]",{},0 (0x0000000000000000),5.209,22.8,4.37704,N/A
107-in_training_update_grad_gamma_beta_gen0487_nhwc_f32,float32,"NHWC [1024, 5, 7, 64]",{},0,41.898000,31.782200,0.758561,0.758561111+in_training_update_grad_gamma_beta_gen0487_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1024, 5, 7, 64], [1024, 5, 7, 64]]",{},0 (0x0000000000000000),41.892,40.5,0.966772,N/A
108-in_training_update_grad_gamma_beta_gen0492_nchw_f32,float32,"NCHW [15, 31, 31, 2]",{},0,4.000000,15.091000,3.772750,3.772750112+in_training_update_grad_gamma_beta_gen0492_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[15, 31, 31, 2], [15, 31, 31, 2]]",{},0 (0x0000000000000000),4.173,15.9,3.810208,N/A
109-in_training_update_grad_gamma_beta_gen0497_nhwc_f32,float32,"NHWC [4, 16, 3, 649]",{},0,4.536000,13.337600,2.940388,2.940388113+in_training_update_grad_gamma_beta_gen0497_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[4, 16, 3, 649], [4, 16, 3, 649]]",{},0 (0x0000000000000000),4.781,15.2,3.179251,N/A
110-in_training_update_grad_gamma_beta_gen0502_nchw_f32,float32,"NCHW [32, 16, 2, 7]",{},0,4.050000,14.924800,3.685136,3.685136114+in_training_update_grad_gamma_beta_gen0502_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[32, 16, 2, 7], [32, 16, 2, 7]]",{},0 (0x0000000000000000),4.384,17.0,3.877737,N/A
111-in_training_update_grad_gamma_beta_gen0507_nhwc_f32,float32,"NHWC [31, 31, 15, 1]",{},0,4.308000,14.553600,3.378273,3.378273115+in_training_update_grad_gamma_beta_gen0507_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[31, 31, 15, 1], [31, 31, 15, 1]]",{},0 (0x0000000000000000),4.491,16.6,3.696281,N/A
112-in_training_update_grad_gamma_beta_gen0512_nchw_f32,float32,"NCHW [257, 8, 31, 2]",{},0,11.206000,24.575800,2.193093,2.193093116+in_training_update_grad_gamma_beta_gen0512_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 8, 31, 2], [257, 8, 31, 2]]",{},0 (0x0000000000000000),11.387,28.0,2.458944,N/A
113-in_training_update_grad_gamma_beta_gen0517_nhwc_f32,float32,"NHWC [63, 31, 2, 64]",{},0,7.436000,24.147000,3.247310,3.247310117+in_training_update_grad_gamma_beta_gen0517_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[63, 31, 2, 64], [63, 31, 2, 64]]",{},0 (0x0000000000000000),7.5,27.5,3.666667,N/A
114-in_training_update_grad_gamma_beta_gen0522_nchw_f32,float32,"NCHW [7, 16, 15, 648]",{},0,15.520000,19.507000,1.256894,1.256894118+in_training_update_grad_gamma_beta_gen0522_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[7, 16, 15, 648], [7, 16, 15, 648]]",{},0 (0x0000000000000000),15.568,22.2,1.426002,N/A
115-in_training_update_grad_gamma_beta_gen0527_nhwc_f32,float32,"NHWC [65, 5, 5, 32]",{},0,5.154000,16.179000,3.139115,3.139115119+in_training_update_grad_gamma_beta_gen0527_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[65, 5, 5, 32], [65, 5, 5, 32]]",{},0 (0x0000000000000000),5.334,18.4,3.449569,N/A
116-in_training_update_grad_gamma_beta_gen0532_nchw_f32,float32,"NCHW [127, 7, 16, 649]",{},0,102.592000,65.740000,0.640791,0.640791120+in_training_update_grad_gamma_beta_gen0532_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[127, 7, 16, 649], [127, 7, 16, 649]]",{},0 (0x0000000000000000),102.67,76.7,0.747054,N/A
117-in_training_update_grad_gamma_beta_gen0537_nhwc_f32,float32,"NHWC [257, 16, 2, 4]",{},0,11.082000,22.240000,2.006858,2.006858121+in_training_update_grad_gamma_beta_gen0537_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[257, 16, 2, 4], [257, 16, 2, 4]]",{},0 (0x0000000000000000),11.062,25.3,2.287109,N/A
118-in_training_update_grad_gamma_beta_gen0542_nchw_f32,float32,"NCHW [513, 7, 5, 64]",{},0,22.524000,38.437800,1.706526,1.706526122+in_training_update_grad_gamma_beta_gen0542_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[513, 7, 5, 64], [513, 7, 5, 64]]",{},0 (0x0000000000000000),22.261,43.7,1.963074,N/A
119-in_training_update_grad_gamma_beta_gen0547_nhwc_f32,float32,"NHWC [5, 1, 2, 63]",{},0,3.312000,9.772800,2.950725,2.950725123+in_training_update_grad_gamma_beta_gen0547_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[5, 1, 2, 63], [5, 1, 2, 63]]",{},0 (0x0000000000000000),3.367,11.2,3.326403,N/A
120-in_training_update_grad_gamma_beta_gen0552_nchw_f32,float32,"NCHW [32, 16, 15, 63]",{},0,9.120000,19.084200,2.092566,2.092566124+in_training_update_grad_gamma_beta_gen0552_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[32, 16, 15, 63], [32, 16, 15, 63]]",{},0 (0x0000000000000000),9.467,21.8,2.302736,N/A
121-in_training_update_grad_gamma_beta_gen0557_nhwc_f32,float32,"NHWC [16, 2, 16, 7]",{},0,3.580000,12.140600,3.391229,3.391229125+in_training_update_grad_gamma_beta_gen0557_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 2, 16, 7], [16, 2, 16, 7]]",{},0 (0x0000000000000000),3.923,13.8,3.517716,N/A
122-in_training_update_grad_gamma_beta_gen0562_nchw_f32,float32,"NCHW [257, 2, 1, 15]",{},0,11.354000,15.225400,1.340972,1.340972126+in_training_update_grad_gamma_beta_gen0562_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 2, 1, 15], [257, 2, 1, 15]]",{},0 (0x0000000000000000),11.462,17.2,1.500611,N/A
123-in_training_update_grad_gamma_beta_gen0567_nhwc_f32,float32,"NHWC [63, 1, 4, 2]",{},0,3.962000,19.525800,4.928269,4.928269127+in_training_update_grad_gamma_beta_gen0567_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[63, 1, 4, 2], [63, 1, 4, 2]]",{},0 (0x0000000000000000),4.051,22.8,5.62824,N/A
124-in_training_update_grad_gamma_beta_gen0572_nchw_f32,float32,"NCHW [31, 5, 31, 257]",{},0,18.174000,19.609400,1.078981,1.078981128+in_training_update_grad_gamma_beta_gen0572_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[31, 5, 31, 257], [31, 5, 31, 257]]",{},0 (0x0000000000000000),18.33,22.4,1.22204,N/A
125-in_training_update_grad_gamma_beta_gen0577_nhwc_f32,float32,"NHWC [513, 5, 8, 8]",{},0,18.930000,36.191600,1.911865,1.911865129+in_training_update_grad_gamma_beta_gen0577_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[513, 5, 8, 8], [513, 5, 8, 8]]",{},0 (0x0000000000000000),19.127,41.4,2.16448,N/A
126-in_training_update_grad_gamma_beta_gen0582_nchw_f32,float32,"NCHW [1024, 4, 3, 5]",{},0,36.240000,24.780800,0.683797,0.683797130+in_training_update_grad_gamma_beta_gen0582_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 4, 3, 5], [1024, 4, 3, 5]]",{},0 (0x0000000000000000),36.726,28.2,0.767848,N/A
127-in_training_update_grad_gamma_beta_gen0587_nhwc_f32,float32,"NHWC [3, 31, 32, 2]",{},0,3.384000,12.851200,3.797636,3.797636131+in_training_update_grad_gamma_beta_gen0587_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 31, 32, 2], [3, 31, 32, 2]]",{},0 (0x0000000000000000),3.583,14.6,4.074798,N/A
128-in_training_update_grad_gamma_beta_gen0592_nchw_f32,float32,"NCHW [2, 2, 3, 647]",{},0,3.448000,12.422400,3.602784,3.602784132+in_training_update_grad_gamma_beta_gen0592_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 2, 3, 647], [2, 2, 3, 647]]",{},0 (0x0000000000000000),3.321,14.1,4.245709,N/A
129-in_training_update_grad_gamma_beta_gen0597_nhwc_f32,float32,"NHWC [15, 2, 32, 16]",{},0,3.634000,15.308800,4.212658,4.212658133+in_training_update_grad_gamma_beta_gen0597_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[15, 2, 32, 16], [15, 2, 32, 16]]",{},0 (0x0000000000000000),3.735,17.5,4.685408,N/A
130-in_training_update_grad_gamma_beta_gen0602_nchw_f32,float32,"NCHW [257, 1, 1, 647]",{},0,11.880000,16.652400,1.401717,1.401717134+in_training_update_grad_gamma_beta_gen0602_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 1, 1, 647], [257, 1, 1, 647]]",{},0 (0x0000000000000000),12.201,19.0,1.557249,N/A
131-in_training_update_grad_gamma_beta_gen0607_nhwc_f32,float32,"NHWC [7, 1, 7, 647]",{},0,4.274000,13.273400,3.105615,3.105615135+in_training_update_grad_gamma_beta_gen0607_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[7, 1, 7, 647], [7, 1, 7, 647]]",{},0 (0x0000000000000000),4.609,15.1,3.276199,N/A
132-in_training_update_grad_gamma_beta_gen0612_nchw_f32,float32,"NCHW [257, 7, 8, 64]",{},0,13.452000,26.213800,1.948692,1.948692136+in_training_update_grad_gamma_beta_gen0612_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 7, 8, 64], [257, 7, 8, 64]]",{},0 (0x0000000000000000),13.733,30.2,2.199083,N/A
133-in_training_update_grad_gamma_beta_gen0617_nhwc_f32,float32,"NHWC [513, 31, 4, 31]",{},0,38.270000,38.041200,0.994021,0.994021137+in_training_update_grad_gamma_beta_gen0617_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[513, 31, 4, 31], [513, 31, 4, 31]]",{},0 (0x0000000000000000),38.052,45.9,1.206244,N/A
134-in_training_update_grad_gamma_beta_gen0622_nchw_f32,float32,"NCHW [7, 7, 2, 8]",{},0,3.332000,10.713600,3.215366,3.215366138+in_training_update_grad_gamma_beta_gen0622_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[7, 7, 2, 8], [7, 7, 2, 8]]",{},0 (0x0000000000000000),3.544,12.3,3.470655,N/A
135-in_training_update_grad_gamma_beta_gen0628_nchw_f32,float32,"NCHW [1, 2, 7, 649]",{},0,3.180000,12.345600,3.882264,3.882264139+in_training_update_grad_gamma_beta_gen0628_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1, 2, 7, 649], [1, 2, 7, 649]]",{},0 (0x0000000000000000),3.257,14.1,4.329137,N/A
136-in_training_update_grad_gamma_beta_gen0633_nhwc_f32,float32,"NHWC [32, 4, 8, 7]",{},0,4.096000,14.976000,3.656250,3.656250140+in_training_update_grad_gamma_beta_gen0633_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[32, 4, 8, 7], [32, 4, 8, 7]]",{},0 (0x0000000000000000),4.165,17.0,4.081633,N/A
137-in_training_update_grad_gamma_beta_gen0638_nchw_f32,float32,"NCHW [2, 31, 32, 7]",{},0,3.334000,12.652600,3.795021,3.795021141+in_training_update_grad_gamma_beta_gen0638_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 31, 32, 7], [2, 31, 32, 7]]",{},0 (0x0000000000000000),3.109,14.5,4.663879,N/A
138-in_training_update_grad_gamma_beta_gen0643_nhwc_f32,float32,"NHWC [257, 16, 15, 1]",{},0,10.868000,24.345400,2.240099,2.240099142+in_training_update_grad_gamma_beta_gen0643_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[257, 16, 15, 1], [257, 16, 15, 1]]",{},0 (0x0000000000000000),10.968,27.8,2.534646,N/A
139-in_training_update_grad_gamma_beta_gen0648_nchw_f32,float32,"NCHW [31, 3, 7, 8]",{},0,4.092000,14.713600,3.595699,3.595699143+in_training_update_grad_gamma_beta_gen0648_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[31, 3, 7, 8], [31, 3, 7, 8]]",{},0 (0x0000000000000000),4.155,16.8,4.043321,N/A
140-in_training_update_grad_gamma_beta_gen0653_nhwc_f32,float32,"NHWC [4, 1, 16, 64]",{},0,3.302000,12.889600,3.903574,3.903574144+in_training_update_grad_gamma_beta_gen0653_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[4, 1, 16, 64], [4, 1, 16, 64]]",{},0 (0x0000000000000000),3.586,14.6,4.071389,N/A
141-in_training_update_grad_gamma_beta_gen0658_nchw_f32,float32,"NCHW [1024, 16, 3, 65]",{},0,47.980000,33.952000,0.707628,0.707628145+in_training_update_grad_gamma_beta_gen0658_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 16, 3, 65], [1024, 16, 3, 65]]",{},0 (0x0000000000000000),48.73,45.9,0.941925,N/A
142-in_training_update_grad_gamma_beta_gen0663_nhwc_f32,float32,"NHWC [3, 16, 2, 3]",{},0,3.072000,10.323200,3.360417,3.360417146+in_training_update_grad_gamma_beta_gen0663_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 16, 2, 3], [3, 16, 2, 3]]",{},0 (0x0000000000000000),3.165,11.8,3.728278,N/A
143-in_training_update_grad_gamma_beta_gen0668_nchw_f32,float32,"NCHW [7, 31, 7, 64]",{},0,4.592000,14.080000,3.066202,3.066202147+in_training_update_grad_gamma_beta_gen0668_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[7, 31, 7, 64], [7, 31, 7, 64]]",{},0 (0x0000000000000000),4.853,16.0,3.29693,N/A
144-in_training_update_grad_gamma_beta_gen0673_nhwc_f32,float32,"NHWC [63, 2, 2, 3]",{},0,4.844000,19.820400,4.091742,4.091742148+in_training_update_grad_gamma_beta_gen0673_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[63, 2, 2, 3], [63, 2, 2, 3]]",{},0 (0x0000000000000000),4.933,22.6,4.581391,N/A
145-in_training_update_grad_gamma_beta_gen0678_nchw_f32,float32,"NCHW [3, 7, 32, 649]",{},0,7.942000,18.060400,2.274037,2.274037149+in_training_update_grad_gamma_beta_gen0678_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[3, 7, 32, 649], [3, 7, 32, 649]]",{},0 (0x0000000000000000),8.25,20.6,2.49697,N/A
146-in_training_update_grad_gamma_beta_gen0683_nhwc_f32,float32,"NHWC [127, 16, 8, 3]",{},0,6.928000,19.142400,2.763048,2.763048150+in_training_update_grad_gamma_beta_gen0683_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[127, 16, 8, 3], [127, 16, 8, 3]]",{},0 (0x0000000000000000),6.96,21.8,3.132184,N/A
147-in_training_update_grad_gamma_beta_gen0688_nchw_f32,float32,"NCHW [63, 16, 7, 3]",{},0,5.044000,20.684600,4.100833,4.100833151+in_training_update_grad_gamma_beta_gen0688_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[63, 16, 7, 3], [63, 16, 7, 3]]",{},0 (0x0000000000000000),5.26,23.5,4.467681,N/A
148-in_training_update_grad_gamma_beta_gen0693_nhwc_f32,float32,"NHWC [513, 1, 2, 648]",{},0,21.340000,36.959600,1.731940,1.731940152+in_training_update_grad_gamma_beta_gen0693_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[513, 1, 2, 648], [513, 1, 2, 648]]",{},0 (0x0000000000000000),21.49,42.3,1.968357,N/A
149-in_training_update_grad_gamma_beta_gen0698_nchw_f32,float32,"NCHW [32, 2, 8, 16]",{},0,4.120000,15.014200,3.644223,3.644223153+in_training_update_grad_gamma_beta_gen0698_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[32, 2, 8, 16], [32, 2, 8, 16]]",{},0 (0x0000000000000000),4.046,17.0,4.201681,N/A
150-in_training_update_grad_gamma_beta_gen0703_nhwc_f32,float32,"NHWC [8, 7, 2, 4]",{},0,3.244000,10.233400,3.154562,3.154562154+in_training_update_grad_gamma_beta_gen0703_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[8, 7, 2, 4], [8, 7, 2, 4]]",{},0 (0x0000000000000000),3.193,12.1,3.78954,N/A
151-in_training_update_grad_gamma_beta_gen0708_nchw_f32,float32,"NCHW [31, 16, 8, 63]",{},0,6.426000,18.937400,2.946997,2.946997155+in_training_update_grad_gamma_beta_gen0708_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[31, 16, 8, 63], [31, 16, 8, 63]]",{},0 (0x0000000000000000),6.568,21.6,3.288672,N/A
152-in_training_update_grad_gamma_beta_gen0713_nhwc_f32,float32,"NHWC [257, 4, 5, 7]",{},0,11.092000,23.551800,2.123314,2.123314156+in_training_update_grad_gamma_beta_gen0713_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[257, 4, 5, 7], [257, 4, 5, 7]]",{},0 (0x0000000000000000),11.311,26.7,2.360534,N/A
153-in_training_update_grad_gamma_beta_gen0718_nchw_f32,float32,"NCHW [64, 4, 1, 64]",{},0,4.910000,15.935600,3.245540,3.245540157+in_training_update_grad_gamma_beta_gen0718_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[64, 4, 1, 64], [64, 4, 1, 64]]",{},0 (0x0000000000000000),5.129,18.1,3.528953,N/A
154-in_training_update_grad_gamma_beta_gen0723_nhwc_f32,float32,"NHWC [127, 31, 1, 31]",{},0,7.492000,33.158000,4.425788,4.425788158+in_training_update_grad_gamma_beta_gen0723_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[127, 31, 1, 31], [127, 31, 1, 31]]",{},0 (0x0000000000000000),7.744,37.9,4.894112,N/A
155-in_training_update_grad_gamma_beta_gen0728_nchw_f32,float32,"NCHW [32, 4, 1, 65]",{},0,4.196000,14.880000,3.546235,3.546235159+in_training_update_grad_gamma_beta_gen0728_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[32, 4, 1, 65], [32, 4, 1, 65]]",{},0 (0x0000000000000000),4.287,16.9,3.942151,N/A
156-in_training_update_grad_gamma_beta_gen0733_nhwc_f32,float32,"NHWC [1024, 4, 32, 15]",{},0,41.462000,31.103400,0.750166,0.750166160+in_training_update_grad_gamma_beta_gen0733_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1024, 4, 32, 15], [1024, 4, 32, 15]]",{},0 (0x0000000000000000),41.526,39.9,0.960844,N/A
157-in_training_update_grad_gamma_beta_gen0738_nchw_f32,float32,"NCHW [1024, 4, 1, 8]",{},0,34.936000,22.418800,0.641711,0.641711161+in_training_update_grad_gamma_beta_gen0738_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 4, 1, 8], [1024, 4, 1, 8]]",{},0 (0x0000000000000000),35.189,25.3,0.718975,N/A
158-in_training_update_grad_gamma_beta_gen0743_nhwc_f32,float32,"NHWC [1024, 2, 4, 64]",{},0,34.386000,26.783400,0.778904,0.778904162+in_training_update_grad_gamma_beta_gen0743_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1024, 2, 4, 64], [1024, 2, 4, 64]]",{},0 (0x0000000000000000),35.112,34.6,0.985418,N/A
159-in_training_update_grad_gamma_beta_gen0748_nchw_f32,float32,"NCHW [2, 3, 3, 649]",{},0,3.440000,12.249600,3.560930,3.560930163+in_training_update_grad_gamma_beta_gen0748_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 3, 3, 649], [2, 3, 3, 649]]",{},0 (0x0000000000000000),3.668,14.0,3.816794,N/A
160-in_training_update_grad_gamma_beta_gen0753_nhwc_f32,float32,"NHWC [15, 7, 1, 649]",{},0,4.664000,14.790000,3.171098,3.171098164+in_training_update_grad_gamma_beta_gen0753_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[15, 7, 1, 649], [15, 7, 1, 649]]",{},0 (0x0000000000000000),4.74,16.8,3.544304,N/A
161-in_training_update_grad_gamma_beta_gen0758_nchw_f32,float32,"NCHW [8, 1, 4, 649]",{},0,3.958000,13.766000,3.478019,3.478019165+in_training_update_grad_gamma_beta_gen0758_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[8, 1, 4, 649], [8, 1, 4, 649]]",{},0 (0x0000000000000000),4.202,15.6,3.712518,N/A
162-in_training_update_grad_gamma_beta_gen0763_nhwc_f32,float32,"NHWC [513, 31, 4, 127]",{},0,99.658000,79.058400,0.793297,0.793297166+in_training_update_grad_gamma_beta_gen0763_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[513, 31, 4, 127], [513, 31, 4, 127]]",{},0 (0x0000000000000000),101.129,86.1,0.851388,N/A
163-in_training_update_grad_gamma_beta_gen0768_nchw_f32,float32,"NCHW [65, 5, 5, 63]",{},0,5.406000,23.417200,4.331706,4.331706167+in_training_update_grad_gamma_beta_gen0768_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[65, 5, 5, 63], [65, 5, 5, 63]]",{},0 (0x0000000000000000),5.643,27.0,4.784689,N/A
164-in_training_update_grad_gamma_beta_gen0773_nhwc_f32,float32,"NHWC [2, 5, 4, 7]",{},0,2.738000,9.836800,3.592695,3.592695168+in_training_update_grad_gamma_beta_gen0773_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[2, 5, 4, 7], [2, 5, 4, 7]]",{},0 (0x0000000000000000),2.722,11.2,4.114622,N/A
165-in_training_update_grad_gamma_beta_gen0778_nchw_f32,float32,"NCHW [64, 5, 4, 649]",{},0,13.374000,18.438400,1.378675,1.378675169+in_training_update_grad_gamma_beta_gen0778_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[64, 5, 4, 649], [64, 5, 4, 649]]",{},0 (0x0000000000000000),13.625,21.1,1.548624,N/A
166-in_training_update_grad_gamma_beta_gen0783_nhwc_f32,float32,"NHWC [31, 16, 16, 65]",{},0,9.222000,19.372400,2.100672,2.100672170+in_training_update_grad_gamma_beta_gen0783_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[31, 16, 16, 65], [31, 16, 16, 65]]",{},0 (0x0000000000000000),9.31,22.1,2.373792,N/A
167-in_training_update_grad_gamma_beta_gen0788_nchw_f32,float32,"NCHW [3, 16, 16, 31]",{},0,3.856000,12.870200,3.337707,3.337707171+in_training_update_grad_gamma_beta_gen0788_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[3, 16, 16, 31], [3, 16, 16, 31]]",{},0 (0x0000000000000000),3.995,14.7,3.679599,N/A
168-in_training_update_grad_gamma_beta_gen0793_nhwc_f32,float32,"NHWC [1, 4, 2, 647]",{},0,3.084000,12.575600,4.077691,4.077691172+in_training_update_grad_gamma_beta_gen0793_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1, 4, 2, 647], [1, 4, 2, 647]]",{},0 (0x0000000000000000),3.147,14.3,4.54401,N/A
169-in_training_update_grad_gamma_beta_gen0798_nchw_f32,float32,"NCHW [2, 7, 8, 127]",{},0,3.188000,12.691200,3.980928,3.980928173+in_training_update_grad_gamma_beta_gen0798_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 7, 8, 127], [2, 7, 8, 127]]",{},0 (0x0000000000000000),3.369,14.5,4.303948,N/A
170-in_training_update_grad_gamma_beta_gen0803_nhwc_f32,float32,"NHWC [1024, 3, 1, 648]",{},0,43.460000,31.142000,0.716567,0.716567174+in_training_update_grad_gamma_beta_gen0803_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1024, 3, 1, 648], [1024, 3, 1, 648]]",{},0 (0x0000000000000000),43.464,39.9,0.918001,N/A
171-in_training_update_grad_gamma_beta_gen0808_nchw_f32,float32,"NCHW [7, 7, 5, 649]",{},0,5.128000,13.682800,2.668253,2.668253175+in_training_update_grad_gamma_beta_gen0808_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[7, 7, 5, 649], [7, 7, 5, 649]]",{},0 (0x0000000000000000),5.406,15.7,2.904181,N/A
172-in_training_update_grad_gamma_beta_gen0813_nhwc_f32,float32,"NHWC [7, 1, 16, 4]",{},0,3.184000,10.707200,3.362814,3.362814176+in_training_update_grad_gamma_beta_gen0813_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[7, 1, 16, 4], [7, 1, 16, 4]]",{},0 (0x0000000000000000),3.184,12.2,3.831658,N/A
173-in_training_update_grad_gamma_beta_gen0818_nchw_f32,float32,"NCHW [16, 4, 32, 16]",{},0,3.874000,15.219000,3.928498,3.928498177+in_training_update_grad_gamma_beta_gen0818_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[16, 4, 32, 16], [16, 4, 32, 16]]",{},0 (0x0000000000000000),4.177,17.4,4.165669,N/A
174-in_training_update_grad_gamma_beta_gen0823_nhwc_f32,float32,"NHWC [16, 7, 4, 127]",{},0,4.300000,15.449400,3.592884,3.592884178+in_training_update_grad_gamma_beta_gen0823_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 7, 4, 127], [16, 7, 4, 127]]",{},0 (0x0000000000000000),4.314,17.6,4.07974,N/A
175-in_training_update_grad_gamma_beta_gen0828_nchw_f32,float32,"NCHW [63, 2, 1, 5]",{},0,4.802000,18.674800,3.888963,3.888963179+in_training_update_grad_gamma_beta_gen0828_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[63, 2, 1, 5], [63, 2, 1, 5]]",{},0 (0x0000000000000000),5.023,21.3,4.240494,N/A
176-in_training_update_grad_gamma_beta_gen0833_nhwc_f32,float32,"NHWC [16, 3, 16, 647]",{},0,8.760000,16.300400,1.860776,1.860776180+in_training_update_grad_gamma_beta_gen0833_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 3, 16, 647], [16, 3, 16, 647]]",{},0 (0x0000000000000000),8.782,18.5,2.106582,N/A
177-in_training_update_grad_gamma_beta_gen0838_nchw_f32,float32,"NCHW [31, 4, 5, 64]",{},0,4.392000,18.297600,4.166120,4.166120181+in_training_update_grad_gamma_beta_gen0838_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[31, 4, 5, 64], [31, 4, 5, 64]]",{},0 (0x0000000000000000),4.529,20.7,4.570545,N/A
178-in_training_update_grad_gamma_beta_gen0843_nhwc_f32,float32,"NHWC [257, 2, 1, 647]",{},0,12.448000,19.558400,1.571208,1.571208182+in_training_update_grad_gamma_beta_gen0843_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[257, 2, 1, 647], [257, 2, 1, 647]]",{},0 (0x0000000000000000),12.632,22.4,1.773274,N/A
179-in_training_update_grad_gamma_beta_gen0848_nchw_f32,float32,"NCHW [63, 7, 5, 649]",{},0,20.974000,24.748600,1.179966,1.179966183+in_training_update_grad_gamma_beta_gen0848_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[63, 7, 5, 649], [63, 7, 5, 649]]",{},0 (0x0000000000000000),21.273,28.6,1.344427,N/A
180-in_training_update_grad_gamma_beta_gen0853_nhwc_f32,float32,"NHWC [2, 3, 4, 15]",{},0,2.804000,9.497400,3.387090,3.387090184+in_training_update_grad_gamma_beta_gen0853_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[2, 3, 4, 15], [2, 3, 4, 15]]",{},0 (0x0000000000000000),2.736,11.2,4.093567,N/A
181-in_training_update_grad_gamma_beta_gen0858_nchw_f32,float32,"NCHW [8, 1, 2, 8]",{},0,2.792000,10.406400,3.727221,3.727221185+in_training_update_grad_gamma_beta_gen0858_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[8, 1, 2, 8], [8, 1, 2, 8]]",{},0 (0x0000000000000000),2.925,12.4,4.239316,N/A
182-in_training_update_grad_gamma_beta_gen0863_nhwc_f32,float32,"NHWC [32, 7, 1, 7]",{},0,4.176000,13.132800,3.144828,3.144828186+in_training_update_grad_gamma_beta_gen0863_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[32, 7, 1, 7], [32, 7, 1, 7]]",{},0 (0x0000000000000000),4.239,15.0,3.53857,N/A
183-in_training_update_grad_gamma_beta_gen0868_nchw_f32,float32,"NCHW [1, 7, 7, 257]",{},0,3.334000,12.236400,3.670186,3.670186187+in_training_update_grad_gamma_beta_gen0868_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1, 7, 7, 257], [1, 7, 7, 257]]",{},0 (0x0000000000000000),3.333,13.9,4.170417,N/A
184-in_training_update_grad_gamma_beta_gen0873_nhwc_f32,float32,"NHWC [2, 7, 4, 31]",{},0,2.936000,12.537200,4.270163,4.270163188+in_training_update_grad_gamma_beta_gen0873_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[2, 7, 4, 31], [2, 7, 4, 31]]",{},0 (0x0000000000000000),3.033,14.3,4.714804,N/A
185-in_training_update_grad_gamma_beta_gen0879_nhwc_f32,float32,"NHWC [15, 2, 31, 648]",{},0,10.262000,16.774200,1.634594,1.634594189+in_training_update_grad_gamma_beta_gen0879_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[15, 2, 31, 648], [15, 2, 31, 648]]",{},0 (0x0000000000000000),10.409,19.1,1.834951,N/A
186-in_training_update_grad_gamma_beta_gen0884_nchw_f32,float32,"NCHW [1, 7, 7, 648]",{},0,3.458000,12.851000,3.716310,3.716310190+in_training_update_grad_gamma_beta_gen0884_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1, 7, 7, 648], [1, 7, 7, 648]]",{},0 (0x0000000000000000),3.589,14.6,4.067986,N/A
187-in_training_update_grad_gamma_beta_gen0889_nhwc_f32,float32,"NHWC [15, 1, 31, 647]",{},0,6.696000,15.449400,2.307258,2.307258191+in_training_update_grad_gamma_beta_gen0889_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[15, 1, 31, 647], [15, 1, 31, 647]]",{},0 (0x0000000000000000),7.422,17.7,2.384802,N/A
188-in_training_update_grad_gamma_beta_gen0894_nchw_f32,float32,"NCHW [5, 2, 8, 64]",{},0,3.382000,13.324800,3.939917,3.939917192+in_training_update_grad_gamma_beta_gen0894_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[5, 2, 8, 64], [5, 2, 8, 64]]",{},0 (0x0000000000000000),3.409,15.2,4.458786,N/A
189-in_training_update_grad_gamma_beta_gen0899_nhwc_f32,float32,"NHWC [7, 5, 15, 5]",{},0,3.418000,10.092800,2.952838,2.952838193+in_training_update_grad_gamma_beta_gen0899_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[7, 5, 15, 5], [7, 5, 15, 5]]",{},0 (0x0000000000000000),3.536,11.4,3.223982,N/A
190-in_training_update_grad_gamma_beta_gen0904_nchw_f32,float32,"NCHW [257, 3, 3, 127]",{},0,12.412000,16.908600,1.362278,1.362278194+in_training_update_grad_gamma_beta_gen0904_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 3, 3, 127], [257, 3, 3, 127]]",{},0 (0x0000000000000000),12.429,19.3,1.55282,N/A
191-in_training_update_grad_gamma_beta_gen0909_nhwc_f32,float32,"NHWC [16, 3, 15, 647]",{},0,8.676000,15.436800,1.779253,1.779253195+in_training_update_grad_gamma_beta_gen0909_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[16, 3, 15, 647], [16, 3, 15, 647]]",{},0 (0x0000000000000000),8.892,17.6,1.979307,N/A
192-in_training_update_grad_gamma_beta_gen0914_nchw_f32,float32,"NCHW [2, 8, 15, 5]",{},0,2.942000,12.466800,4.237525,4.237525196+in_training_update_grad_gamma_beta_gen0914_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[2, 8, 15, 5], [2, 8, 15, 5]]",{},0 (0x0000000000000000),3.023,14.2,4.697321,N/A
193-in_training_update_grad_gamma_beta_gen0919_nhwc_f32,float32,"NHWC [4, 8, 8, 15]",{},0,3.302000,12.870400,3.897759,3.897759197+in_training_update_grad_gamma_beta_gen0919_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[4, 8, 8, 15], [4, 8, 8, 15]]",{},0 (0x0000000000000000),3.486,14.6,4.188181,N/A
194-in_training_update_grad_gamma_beta_gen0924_nchw_f32,float32,"NCHW [15, 7, 4, 31]",{},0,4.002000,15.001400,3.748476,3.748476198+in_training_update_grad_gamma_beta_gen0924_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[15, 7, 4, 31], [15, 7, 4, 31]]",{},0 (0x0000000000000000),4.12,17.1,4.150485,N/A
195-in_training_update_grad_gamma_beta_gen0929_nhwc_f32,float32,"NHWC [15, 3, 16, 3]",{},0,3.636000,12.044800,3.312651,3.312651199+in_training_update_grad_gamma_beta_gen0929_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[15, 3, 16, 3], [15, 3, 16, 3]]",{},0 (0x0000000000000000),3.746,13.7,3.657234,N/A
196-in_training_update_grad_gamma_beta_gen0934_nchw_f32,float32,"NCHW [257, 1, 8, 65]",{},0,11.064000,24.371200,2.202748,2.202748200+in_training_update_grad_gamma_beta_gen0934_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 1, 8, 65], [257, 1, 8, 65]]",{},0 (0x0000000000000000),11.495,27.8,2.418443,N/A
197-in_training_update_grad_gamma_beta_gen0939_nhwc_f32,float32,"NHWC [4, 8, 8, 4]",{},0,3.094000,9.907200,3.202069,3.202069201+in_training_update_grad_gamma_beta_gen0939_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[4, 8, 8, 4], [4, 8, 8, 4]]",{},0 (0x0000000000000000),3.297,11.3,3.427358,N/A
198-in_training_update_grad_gamma_beta_gen0944_nchw_f32,float32,"NCHW [15, 2, 5, 32]",{},0,3.598000,12.307200,3.420567,3.420567202+in_training_update_grad_gamma_beta_gen0944_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[15, 2, 5, 32], [15, 2, 5, 32]]",{},0 (0x0000000000000000),3.753,14.0,3.730349,N/A
199-in_training_update_grad_gamma_beta_gen0949_nhwc_f32,float32,"NHWC [513, 16, 2, 65]",{},0,23.470000,37.849200,1.612663,1.612663203+in_training_update_grad_gamma_beta_gen0949_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[513, 16, 2, 65], [513, 16, 2, 65]]",{},0 (0x0000000000000000),24.021,43.0,1.7901,N/A
200-in_training_update_grad_gamma_beta_gen0954_nchw_f32,float32,"NCHW [4, 16, 31, 31]",{},0,4.232000,13.152000,3.107750,3.107750204+in_training_update_grad_gamma_beta_gen0954_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[4, 16, 31, 31], [4, 16, 31, 31]]",{},0 (0x0000000000000000),4.478,15.0,3.34971,N/A
201-in_training_update_grad_gamma_beta_gen0959_nhwc_f32,float32,"NHWC [1, 7, 7, 649]",{},0,3.536000,12.364800,3.496833,3.496833205+in_training_update_grad_gamma_beta_gen0959_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1, 7, 7, 649], [1, 7, 7, 649]]",{},0 (0x0000000000000000),3.398,14.1,4.1495,N/A
202-in_training_update_grad_gamma_beta_gen0964_nchw_f32,float32,"NCHW [1024, 8, 1, 3]",{},0,32.916000,18.912000,0.574553,0.574553206+in_training_update_grad_gamma_beta_gen0964_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[1024, 8, 1, 3], [1024, 8, 1, 3]]",{},0 (0x0000000000000000),34.24,18.5,0.540304,N/A
203-in_training_update_grad_gamma_beta_gen0969_nhwc_f32,float32,"NHWC [64, 3, 31, 64]",{},0,7.846000,16.825200,2.144430,2.144430207+in_training_update_grad_gamma_beta_gen0969_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[64, 3, 31, 64], [64, 3, 31, 64]]",{},0 (0x0000000000000000),8.088,19.3,2.386251,N/A
204-in_training_update_grad_gamma_beta_gen0974_nchw_f32,float32,"NCHW [513, 8, 5, 4]",{},0,19.052000,33.996400,1.784401,1.784401208+in_training_update_grad_gamma_beta_gen0974_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[513, 8, 5, 4], [513, 8, 5, 4]]",{},0 (0x0000000000000000),19.385,38.8,2.001548,N/A
205-in_training_update_grad_gamma_beta_gen0979_nhwc_f32,float32,"NHWC [1024, 2, 8, 7]",{},0,33.960000,34.892400,1.027456,1.027456209+in_training_update_grad_gamma_beta_gen0979_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1024, 2, 8, 7], [1024, 2, 8, 7]]",{},0 (0x0000000000000000),34.815,39.9,1.146058,N/A
206-in_training_update_grad_gamma_beta_gen0984_nchw_f32,float32,"NCHW [64, 2, 8, 3]",{},0,4.928000,19.635200,3.984416,3.984416210+in_training_update_grad_gamma_beta_gen0984_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[64, 2, 8, 3], [64, 2, 8, 3]]",{},0 (0x0000000000000000),5.25,22.3,4.247619,N/A
207-in_training_update_grad_gamma_beta_gen0989_nhwc_f32,float32,"NHWC [3, 2, 7, 64]",{},0,3.316000,12.665400,3.819481,3.819481211+in_training_update_grad_gamma_beta_gen0989_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[3, 2, 7, 64], [3, 2, 7, 64]]",{},0 (0x0000000000000000),3.459,14.4,4.163053,N/A
208-in_training_update_grad_gamma_beta_gen0994_nchw_f32,float32,"NCHW [257, 1, 7, 2]",{},0,10.560000,14.611200,1.383636,1.383636212+in_training_update_grad_gamma_beta_gen0994_nchw_f32,"['float32', 'float32']","['NCHW', 'NCHW'] [[257, 1, 7, 2], [257, 1, 7, 2]]",{},0 (0x0000000000000000),10.882,16.6,1.525455,N/A
209-in_training_update_grad_gamma_beta_gen0999_nhwc_f32,float32,"NHWC [1, 5, 7, 7]",{},0,2.376000,11.603000,4.883418,4.883418213+in_training_update_grad_gamma_beta_gen0999_nhwc_f32,"['float32', 'float32']","['NHWC', 'NHWC'] [[1, 5, 7, 7], [1, 5, 7, 7]]",{},0 (0x0000000000000000),2.424,13.1,5.40429,N/A
Dchangwei_deliverables/september/INTrainingUpdateGradGammaBeta/HANDOFF_MANIFEST.md+0-85
@@ -1,85 +0,0 @@
1-# INTrainingUpdateGradGammaBeta Ascend 950 转测交付件清单
2- 
3-## 交付结论
4- 
5-- 交付对象：Cannbot 当前 Ascend 950（DAV_3510 / arch35）实现。
6-- canndev 公共契约兼容结论：**PASS**。输入输出数量与顺序、FP32、4D NCHW/NHWC、
7- `sum(axis=0, keepdims=True)`、无属性和 GEIR 调用语义均一致。
8-- 当前实现还支持 NCDHW、NDHWC、4D/5D ND、rank 5 和空 Tensor；这些是 Ascend 950 扩展能力，
9- 不改变 canndev 公共契约。
10-- `NDC1HWC0`/`NC1HWC0`、ACLNN、Torch、TensorFlow、ONNX、非连续或别名 Tensor 不在本次
11- 对外范围，交付件没有宣称支持。
12-- ops-nn PR：<https://gitcode.com/cann/ops-nn/merge_requests/10488>
13- - 本地代码：`/workspace/cannbot-dev/INTrainingUpdateGradGammaBeta/ops-nn`
14- - 分支：`feat/in-training-update-grad-gamma-beta-cannbot-ascend950`
15- - HEAD：`1fe720b94fff1d9a9cbae12f2a9cc44a4a2e0936`
16- - 本地 HEAD 与远端同名分支一致，工作树干净，当前算子修改已全部推送。
17- - 清单整理时 `upstream/master` 为 `b2233739178425367fe73b42d802247beeb836d6`；
18- 当前分支相对其 0 落后并仅含 1 个算子提交。rebase 前后算子目录内容完全一致。
19- 
20-## 文件清单
21- 
22-| 类别 | 路径 | 内容 |
23-|---|---|---|
24-| 交付检查 | `00_check_findings.md` | 自动扫描提示、人工判定及未伪造闭环项 |
25-| 需求 | `01_requirement.md` | 功能、接口、约束、范围和验收标准 |
26-| 设计 | `02_design.md` | 当前 RegBase Kernel、Host Tiling、分支策略和验证结果 |
27-| 规格 | `03_spec.yaml` | 当前 IR/REG_OP 唯一规格 |
28-| Golden | `04_golden/GOLDEN_REF.md` | CPU Golden、Torch-A100 第三方参考和精度判据 |
29-| 泛化结果 | `05_test_result/generalization/INTrainingUpdateGradGammaBeta_generalization.csv` | 1000 条公共 4D + 29 条全格式/边界定向回归，合计 1029/1029 PASS |
30-| GEIR 通路 | `05_test_result/invoke_path/INTrainingUpdateGradGammaBeta_invoke_path.csv` | 合并环境异常复验后的 29 条接受集，29/29 PASS |
31-| 性能结果 | `05_test_result/performance/INTrainingUpdateGradGammaBeta_performance.csv` | 同一批 200 条公共 4D shape 的 A5/A100 数据，逐例 `GPU/A5 > 0.1` |
32-| 三方精度 | `05_test_result/precision/INTrainingUpdateGradGammaBeta_new_precision_3way.csv` | 100/100 case、两个输出 200/200 指标 PASS |
33-| 算子信息 | `06_operator_info.xlsx` | 转测算子信息表 |
34-| 自验证 | `07_self_verification.xlsx` | 转测自验证报告及不适用项依据 |
35- 
36-## 已完成验证
37- 
38-- 最终提交配置归属回归：op_graph 4/4、Ascend 950 op_host 38/38、无 SoC 聚合
39- op_host 18/18 PASS。
40-- CANN 9.2 与任务私有 CANN 9.0.0 均完成 `--ccache=off` clean 定向构建，生成 5 组
41- Kernel object/JSON；构建并发均为 `-j8`。
42-- Golden pytest：7/7 PASS。
43-- 当前 release binary Kernel：新增公共 4D 集 1000/1000 PASS，原定向集 29/29 PASS，
44- 合计 1029 个唯一用例，精度与 OOB 检查全部通过。
45-- GEIR：首批 28/29；唯一失败为并发环境下 `GEInitialize failed`，必要复验 1/1 PASS，
46- 接受集合并后 29/29 PASS。
47-- 动态 OOM 插桩 safety：29/29 PASS。
48-- 全格式三方回归：29/29 case、58/58 输出指标 PASS。
49-- 中等 shape 三方精度：100/100 case、Torch provider 100/100、两个输出 200/200
50- `cross_check` PASS；首轮 10 条 GPU 网关心跳假阴性仅做必要复验并全部通过。
51-- 同一 200 条公共 4D shape 性能：A5 与 A100 各 200/200 数值有效；
52- `GPU/A5 > 0.1` 200/200 PASS，最小 0.554458、p50 3.205499；A5 在 178 条上更快，
53- A100 在 22 条上更快。A100 口径为两次原生 `torch.sum` 的 eager/composite
54- 基线，不宣称为 fused/native 单算子。
55-- 数值证据在提交 `6e6c9407...` 上采集。最终提交 `1fe720b9...` 没有修改 Tiling、Kernel
56- 计算体、Golden 或测试输入；只调整配置/注册归属、文档和 Kernel 入口文件名，故没有重跑
57- 精度/性能全量集。最终提交的定向 UT 与双版本 CANN clean 构建已通过。
58- 
59-## 覆盖度说明
60- 
61-- 当前真实泛化明细为 1029 条：1000 条公共 4D FP32 NCHW/NHWC 与 29 条全格式/边界
62- 定向回归，身份无重复；已满足通用 full 泛化数量门槛。
63-- 当前 GEIR 明细为 29 条，低于通用每通路 50 条门槛；若接收方将该门槛作为硬性要求，需补至少
64- 21 条 GEIR 用例。
65-- 当前 200 条 A5/A100 性能属于 Kernel 档，不是 GEIR 同次、同通路 A100 性能，不能拼接成该证据。
66-- 以上是严格通用交付规则下的覆盖量差异，不表示现有用例失败，也不是已发现的 canndev 公共契约
67- 兼容缺陷。完整人工判定见 `00_check_findings.md`。
68- 
69-## 完整性校验
70- 
71-以下 SHA-256 不包含本清单自身。
72- 
73-| SHA-256 | 相对路径 |
74-|---|---|
75-| `f1b30c306c18a63278d778f8d5b1c10ddbb08889af6fc5235f7cadd2ea201912` | `00_check_findings.md` |
76-| `3213fd5ab1565f305e7e8c1ffd8b88857ec2f8ab4f0eedcac2a85bf8d0c1225e` | `01_requirement.md` |
77-| `4660d077034bc9ad748b7bc41cf3e059cc05b3ed5ad5a754bc44e6473a3e2cd2` | `02_design.md` |
78-| `1eaeabea3ea1b5d94e4202c31a5d501acdf44c95bd6a998986f8ac400d21fd9d` | `03_spec.yaml` |
79-| `f02222d99d5772c40a4e194b25c129ad8385fbe9d257149a35dfbd0451994b69` | `04_golden/GOLDEN_REF.md` |
80-| `d28827ec5f34a730f721a9926f8a470a18dee97b29d506067e5cead34b58e455` | `05_test_result/generalization/INTrainingUpdateGradGammaBeta_generalization.csv` |
81-| `5f5fbd09596aeaa2fb9f9d44d5cb25b759b758c3be8febc2d08a20c7539be059` | `05_test_result/invoke_path/INTrainingUpdateGradGammaBeta_invoke_path.csv` |
82-| `903b094f3c4ba1448e44a45b87e96234052624e1a34e3abd397e79f33c84fbc5` | `05_test_result/performance/INTrainingUpdateGradGammaBeta_performance.csv` |
83-| `50cd42246861d8fe53b1ee1f188fac4e4df3f81a0b1f14590230b06f5efe03a0` | `05_test_result/precision/INTrainingUpdateGradGammaBeta_new_precision_3way.csv` |
84-| `8633df0dc222f1ce0919e11b4dd228b06e78707314d51d6eae7c3ea4f1aaa38d` | `06_operator_info.xlsx` |
85-| `d65b2aee9ce178f1f5726dcbf9103ad1f31495ec0b045d10db952d3bc299d4df` | `07_self_verification.xlsx` |
Achangwei_deliverables/september/INTrainingUpdateGradGammaBeta/codecheck_closure.md+13-0
@@ -0,0 +1,13 @@
1+# 交付整理与残留项
2+ 
3+| 项目 | 结果 | 依据 |
4+|---|---|---|
5+| 标准资产 | 01–07齐备；05四类各一份说明+明细CSV | 对照标杆项目与现行单文件规范 |
6+| 逐输出精度 | 验收2152输出原8列不变，补原始rmse/判定/分类 | 固定版本source_map原始指标 |
7+| 原始证据 | 完整旧交付副本SHA一致后迁出标准目录 | 固定Git提交及本地持久化archive |
8+| 接口范围 | 仅Kernel直调、GEIR；ACLNN/Torch接口不支持 | 01/02/06/07一致；Torch-A100仅参考 |
9+| 正式验收范围 | 1076/1076 PASS，2152输出指标原值保留 | [独立验收范围清单](https://gitcode.com/zl_hw/myOpen/tree/49dc5703162b7e5ad146d32adbb1665984eb0242/handover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076)；排除记录未改判 |
10+| GEIR三方精度 | 50/50用例、100/100输出cross_check PASS，GPU/NPU最小0.842386 | 同次GPU/SQLite计时、check_invoke_path退出0；旧两方批次独立留档 |
11+| 生产代码全量审查 | 不在本次交付整理结论内 | 仅执行组15及交付检查，不复制标杆代码结论 |
12+ 
13+“格式通过”不等于“真实三方完整”。本文件不作全量开发门禁放行。
Achangwei_deliverables/september/INTrainingUpdateGradGammaBeta/codecheck_report.md+17-0
@@ -0,0 +1,17 @@
1+# 交付件检查报告
2+ 
3+## 检查范围
4+ 
5+本报告仅检查INTrainingUpdateGradGammaBeta交付件。实际执行changwei-code-check的check.py组15、audit_deliverables.py、scan_handover.py，以及原始GEIR结果的check_invoke_path.py。组15扫描定义域包含59个源文件，但仅交付格式规则执行；不代表生产代码完整审查、多SOC编译、UT、OAT或完整CI通过。
6+ 
7+## 实际结果
8+ 
9+- audit_deliverables.py：标准文件和四类单CSV格式通过。
10+- check.py组15：0条格式问题；不能据此宣称三方证据完整。
11+- scan_handover.py：2条低风险提示；06含5个不支持接口“不涉及”，07含7个“不涉及”（不支持接口、范围外融合和非连续Tensor），均有依据保留。
12+- 当前GEIR50例、100输出真实cross_check PASS；同次GPU与SQLite计时最小GPU/NPU=0.842386，50例均>0.1。check_invoke_path退出0。原始CST_OFF保留，不是TTK原生性能PASS。旧stat_rel_err批次仅为历史，不混入当前50例。
13+- Kernel正式验收1076例、2152输出PASS；仅精确裁剪用户指定的两个GPU参考溢出用例，真实指标未改写。原FAIL和完整旧表见[独立验收范围清单](https://gitcode.com/zl_hw/myOpen/tree/49dc5703162b7e5ad146d32adbb1665984eb0242/handover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076)。
14+ 
15+## 放行边界
16+ 
17+GEIR同次真实三方精度与计时门禁通过。私有CANN二进制补丁不等同正式canndev源码构建。仅执行交付检查与组15，不宣称生产源码全量审查、多SOC构建或完整CI通过。
Ahandover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076/README.md+9-0
@@ -0,0 +1,9 @@
1+# 正式验收集范围
2+ 
3+用户明确授权正式验收集调整为1076条，按scope_manifest.json中两个精确testcase ID裁剪；不按FAIL批量过滤，不改变任何保留行指标、阈值或判定。
4+ 
5+正式精度集为1076例、2152输出，全部PASS；泛化集采用同一1076例范围。完整原精度表1078例、2156输出中1076例PASS、2例FAIL，未被改判。原两方泛化1078例全部PASS亦原样保留。
6+ 
7+两个排除用例是NCHW overflow_boundary_equal与NHWC overflow_boundary_below（完整ID见清单）。GPU torch.sum FP32参考累加溢出，原三方cross_check记录NaN/Inf mismatch及双输出FAIL；NPU对CPU高精度Golden的独立stat_rel_err双输出PASS，release binary均BINARY_MATCH。范围裁剪不表示GPU行为被修复，不表示原1078三方全部通过。
8+ 
9+本目录保留裁剪前完整两张标准表及三方/两方原始5例结果CSV，字节SHA和原始来源commit见scope_manifest.json。更完整的raw、日志及来源映射仍可从清单中的历史Git祖先commit读取，未删除或覆盖。本范围调整未执行设备测试或构建。GEIR50和Kernel200性能表不受影响。
Ahandover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076/scope_manifest.json+59-0
@@ -0,0 +1,59 @@
1+{
2+ "authorization": "用户明确要求正式验收集调整为1076，并从正式表精确排除以下两个GPU参考溢出用例；不得抹去失败证据或宣称原1078全部通过。",
3+ "rule": "exact testcase ID exclusion only; no threshold, metric or verdict changes",
4+ "excluded_case_ids": [
5+ "in_training_update_grad_gamma_beta_nchw_overflow_boundary_equal",
6+ "in_training_update_grad_gamma_beta_nhwc_overflow_boundary_below"
7+ ],
8+ "original_three_way": "2 cases / 4 outputs FAIL; cross_check NaN/Inf mismatch; GPU FP32 reference accumulation overflow",
9+ "original_two_way": "2 cases / 4 outputs PASS; NPU vs CPU FP64-promoted Golden stat_rel_err; BINARY_MATCH",
10+ "before": {
11+ "cases": 1078,
12+ "outputs": 2156,
13+ "pass_cases": 1076,
14+ "fail_cases": 2
15+ },
16+ "formal_acceptance": {
17+ "cases": 1076,
18+ "outputs": 2152,
19+ "pass_cases": 1076,
20+ "fail_cases": 0
21+ },
22+ "generalization": {
23+ "before": 1078,
24+ "after": 1076,
25+ "original_two_way_all_pass": true
26+ },
27+ "historical_evidence_commit": {
28+ "myOpen": "471e13bb05bf992586d6799ce310af4aff311089",
29+ "cannbot-deliverables": "cfbcd2efe374fcc8add51f0672776f627ecc0e7a"
30+ },
31+ "source_standard_commit": {
32+ "myOpen": "7561310baf5ef415d98cd137dfa4460c7c5f8046",
33+ "cannbot-deliverables": "d7b0c0ed327fe7486939844e5fb33d75b2854d63"
34+ },
35+ "preserved_files": {
36+ "precision_before_1078.csv": "d4e3bc59217c0efc1274099c240b7532394271d027a199db75f789d135c51303",
37+ "generalization_before_1078.csv": "488641be08586de65d6731714ef54dcc13782b461bfbbfd4a633247ca9737b9f",
38+ "three_way_result.csv": "90793ccb7606964459e39fa2841d1c5cb6e2fdaf8746af5ab764380cb9b3a0fc",
39+ "two_way_result.csv": "b268bc9c243944d6b4c6e633f9320fb03c94a48e4c86bfc247ee7bb2f221496b"
40+ },
41+ "standard_before_sha256": {
42+ "01_requirement.md": "7420d912a11baa5392d60cf6e821ee8283ba66d5448a79136509212a52d56e81",
43+ "02_design.md": "c3d51ca70bd5793bac724bf6573f1a8b1f790d1979b600219d1c663d6d0faa24",
44+ "03_spec.yaml": "6e53f82cf233356171ce4ac64bc2fe8a48e4c69d368d836c779f5533e084c229",
45+ "00_check_findings.md": "44e295e3d51d42f467bb293233d2ecd203679612d987a0980745ad0659ddad29",
46+ "06_operator_info.xlsx": "74efb0810eb7c1709d2f1bb33d435e9b516fd15d62b3ece2c3230bd2805ad8e1",
47+ "07_self_verification.xlsx": "63b8a497d0fd02dc7c9932434c7d33ffe3e0319df06a435906b1ab3858ee69e5",
48+ ".gitattributes": "8f21015b144cb8d95db1ba15ee63bc8b30e3971c57ad2498d746030c79e667a6",
49+ "codecheck_report.md": "f4ee3ce9eeb27de39bb3100fffb7c87ced9116769ead55bd4d796c26cc7bf872",
50+ "codecheck_closure.md": "a97f4c095870658179d978bba9391ae67838cca0d8349261ce0f1e50b90ec51a",
51+ "04_golden/GOLDEN_REF.md": "171c4d76cf8b81a715c68da18600e7aecbee8d5598d9997f7285fad4f10d1c0e",
52+ "05_test_result/generalization/INTrainingUpdateGradGammaBeta_generalization.csv": "488641be08586de65d6731714ef54dcc13782b461bfbbfd4a633247ca9737b9f",
53+ "05_test_result/invoke_path/INTrainingUpdateGradGammaBeta_invoke_path.csv": "47abaedb8b9ddff4a4be96166e09d754be13f301745558005fa0b32c51a25dce",
54+ "05_test_result/performance/INTrainingUpdateGradGammaBeta_performance.csv": "a067ba69d7214e7e911afeca629752229f120a833567ed7966f876e0623d58f1",
55+ "05_test_result/precision/INTrainingUpdateGradGammaBeta_new_precision_3way.csv": "d4e3bc59217c0efc1274099c240b7532394271d027a199db75f789d135c51303"
56+ },
57+ "no_device_rerun": true,
58+ "note": "This scope decision does not fix or reclassify the GPU behavior. Full historical raw evidence remains reachable in ancestor commits."
59+}
Ahandover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076/three_way_result.csv+6-0
@@ -0,0 +1,6 @@
1+testcase_name,op_name,dyn_tiling_time_us,bin_tiling_time_us,dyn_kernel_name,cst_kernel_name,bin_kernel_name,dyn_block_dim,cst_block_dim,bin_block_dim,dyn_perf_us,cst_perf_us,bin_perf_us,dyn_compile_s,cst_compile_s,bin_compile_s,perf_status,dyn_precision,cst_precision,bin_precision,precision_status,dyn_oob_result,cst_oob_result,bin_oob_result,memory_oob_status,dyn_tiling_key,dyn_tiling_data,bin_tiling_key,bin_tiling_data,dyn_workspaces,cst_workspaces,bin_workspaces,data_input_size_b,data_output_size_b,soc,xpu_metrics,precision_metrics,network_name,input_dtypes,input_ori_shapes,output_ori_shapes,input_ori_formats,output_ori_formats,attributes,attributes1,attributes2,attributes3,attributes4,attributes5,attributes6,attributes7,attributes8,attributes9,input_shapes,output_dtypes,output_shapes,input_formats,output_formats,input_data_ranges,precision_tolerances,absolute_precision,output_inplace_indexes,output_shape_unknown_indexes,is_enabled,remark,soc_series,priority,dump_file_prefix,manual_input_binaries,manual_golden_binaries,manual_xpu_binaries
2+in_training_update_grad_gamma_beta_nd_cancellation_orders,in_training_update_grad_gamma_beta,DYN_OFF,"(11, 9, 9)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,3,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"PASS,PASS",PASS,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(4, 0, 24, 0, 1, 0, 8, 4, 0, 8, 3, 2113929216, 1040187392, 1090519040)",,,"(16777216,)",768,192,Ascend950PR_9579,"{'torch': {'status': 'PASS', 'api': 'InTrainingUpdateGradGammaBetaSpec'}}","{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': 0.0, 'mere': 0.0, 'rmse': 0.0, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': True}, 1: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': 0.0, 'mere': 0.0, 'rmse': 0.0, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': True}}}",UNKNOWN,"('float32', 'float32')","((4, 24, 1, 1), (4, 24, 1, 1))","((1, 24, 1, 1), (1, 24, 1, 1))","('ND', 'ND')","('ND', 'ND')",{},{},{},{},{},{},{},{},{},{},"((4, 24, 1, 1), (4, 24, 1, 1))","('float32', 'float32')","((1, 24, 1, 1), (1, 24, 1, 1))","('ND', 'ND')","('ND', 'ND')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,all 24 permutations of extreme cancellation,,0,,(),(),()
3+in_training_update_grad_gamma_beta_ndhwc_scale_boundary_neighbors,in_training_update_grad_gamma_beta,DYN_OFF,"(10, 9, 15)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,3,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"PASS,PASS",PASS,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(8193, 0, 17, 0, 1, 0, 8, 4095, 0, 8, 3, 2013265920, 939524096, 1191182336)",,,"(16777216,)",1114248,136,Ascend950PR_9579,"{'torch': {'status': 'PASS', 'api': 'InTrainingUpdateGradGammaBetaSpec'}}","{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': 0.0009763241687146319, 'mere': 0.00022976540254174957, 'rmse': 0.5000000111785997, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': True}, 1: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': 0.0009763241687146319, 'mere': 0.00022976540254174957, 'rmse': 0.5000000111785997, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': True}}}",UNKNOWN,"('float32', 'float32')","((8193, 1, 1, 1, 17), (8193, 1, 1, 1, 17))","((1, 1, 1, 1, 17), (1, 1, 1, 1, 17))","('NDHWC', 'NDHWC')","('NDHWC', 'NDHWC')",{},{},{},{},{},{},{},{},{},{},"((8193, 1, 1, 1, 17), (8193, 1, 1, 1, 17))","('float32', 'float32')","((1, 1, 1, 1, 17), (1, 1, 1, 1, 17))","('NDHWC', 'NDHWC')","('NDHWC', 'NDHWC')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,power-of-two row-count neighbor with output tail,,0,,(),(),()
4+in_training_update_grad_gamma_beta_ncdhw_scale_boundary_neighbors,in_training_update_grad_gamma_beta,DYN_OFF,"(21, 32, 13)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,2,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"PASS,PASS",PASS,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(10000, 0, 13, 0, 1, 0, 8, 4095, 0, 8, 2, 2013265920, 939524096, 1191182336)",,,"(16777216,)",1040000,104,Ascend950PR_9579,"{'torch': {'status': 'PASS', 'api': 'InTrainingUpdateGradGammaBetaSpec'}}","{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': 0.000623437537159773, 'mere': 0.00022536057578399944, 'rmse': 1.0, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': True}, 1: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': 0.000623437537159773, 'mere': 0.00022536057578399944, 'rmse': 1.0, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': True}}}",UNKNOWN,"('float32', 'float32')","((10000, 13, 1, 1, 1), (10000, 13, 1, 1, 1))","((1, 13, 1, 1, 1), (1, 13, 1, 1, 1))","('NCDHW', 'NCDHW')","('NCDHW', 'NCDHW')",{},{},{},{},{},{},{},{},{},{},"((10000, 13, 1, 1, 1), (10000, 13, 1, 1, 1))","('float32', 'float32')","((1, 13, 1, 1, 1), (1, 13, 1, 1, 1))","('NCDHW', 'NCDHW')","('NCDHW', 'NCDHW')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,scaled and unscaled columns with one ULP neighbors,,0,,(),(),()
5+in_training_update_grad_gamma_beta_nhwc_overflow_boundary_below,in_training_update_grad_gamma_beta,DYN_OFF,"(14, 13, 12)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,2,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"FAIL,FAIL",FAIL,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(10000, 0, 13, 0, 1, 0, 8, 4095, 0, 8, 2, 2013265920, 939524096, 1191182336)",,,"(16777216,)",1040000,104,Ascend950PR_9579,"{'torch': {'status': 'PASS', 'api': 'InTrainingUpdateGradGammaBetaSpec'}}","{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': None, 'mere': None, 'rmse': None, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': False, 'reason': 'NaN/Inf mismatch'}, 1: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': None, 'mere': None, 'rmse': None, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': False, 'reason': 'NaN/Inf mismatch'}}}",UNKNOWN,"('float32', 'float32')","((10000, 1, 1, 13), (10000, 1, 1, 13))","((1, 1, 1, 13), (1, 1, 1, 13))","('NHWC', 'NHWC')","('NHWC', 'NHWC')",{},{},{},{},{},{},{},{},{},{},"((10000, 1, 1, 13), (10000, 1, 1, 13))","('float32', 'float32')","((1, 1, 1, 13), (1, 1, 1, 13))","('NHWC', 'NHWC')","('NHWC', 'NHWC')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,one ULP below maximum-sum boundary,,0,,(),(),()
6+in_training_update_grad_gamma_beta_nchw_overflow_boundary_equal,in_training_update_grad_gamma_beta,DYN_OFF,"(24, 10, 25)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,2,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"FAIL,FAIL",FAIL,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(10000, 0, 13, 0, 1, 0, 8, 4095, 0, 8, 2, 2013265920, 939524096, 1191182336)",,,"(16777216,)",1040000,104,Ascend950PR_9579,"{'torch': {'status': 'PASS', 'api': 'InTrainingUpdateGradGammaBetaSpec'}}","{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': None, 'mere': None, 'rmse': None, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': False, 'reason': 'NaN/Inf mismatch'}, 1: {'standard': 'cross_check', 'level': 'L1', 'config': {'mare': 5.0, 'mere': 1.5, 'rmse': 1.5, 'small_value': 6.103515625e-05, 'small_value_atol': 9.313225746154785e-10}, 'result': {'mare': None, 'mere': None, 'rmse': None, 'small_err_cnt_target': 0, 'small_err_cnt_third': 0, 'small_err_ratio': 0.0}, 'pass': False, 'reason': 'NaN/Inf mismatch'}}}",UNKNOWN,"('float32', 'float32')","((10000, 13, 1, 1), (10000, 13, 1, 1))","((1, 13, 1, 1), (1, 13, 1, 1))","('NCHW', 'NCHW')","('NCHW', 'NCHW')",{},{},{},{},{},{},{},{},{},{},"((10000, 13, 1, 1), (10000, 13, 1, 1))","('float32', 'float32')","((1, 13, 1, 1), (1, 13, 1, 1))","('NCHW', 'NCHW')","('NCHW', 'NCHW')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,finite sum near FP32 maximum,,0,,(),(),()
Ahandover_evidence/INTrainingUpdateGradGammaBeta/acceptance_scope_1076/two_way_result.csv+6-0
@@ -0,0 +1,6 @@
1+testcase_name,op_name,dyn_tiling_time_us,bin_tiling_time_us,dyn_kernel_name,cst_kernel_name,bin_kernel_name,dyn_block_dim,cst_block_dim,bin_block_dim,dyn_perf_us,cst_perf_us,bin_perf_us,dyn_compile_s,cst_compile_s,bin_compile_s,perf_status,dyn_precision,cst_precision,bin_precision,precision_status,dyn_oob_result,cst_oob_result,bin_oob_result,memory_oob_status,dyn_tiling_key,dyn_tiling_data,bin_tiling_key,bin_tiling_data,dyn_workspaces,cst_workspaces,bin_workspaces,data_input_size_b,data_output_size_b,soc,xpu_metrics,precision_metrics,network_name,input_dtypes,input_ori_shapes,output_ori_shapes,input_ori_formats,output_ori_formats,attributes,attributes1,attributes2,attributes3,attributes4,attributes5,attributes6,attributes7,attributes8,attributes9,input_shapes,output_dtypes,output_shapes,input_formats,output_formats,input_data_ranges,precision_tolerances,absolute_precision,output_inplace_indexes,output_shape_unknown_indexes,is_enabled,remark,soc_series,priority,dump_file_prefix,manual_input_binaries,manual_golden_binaries,manual_xpu_binaries
2+in_training_update_grad_gamma_beta_nd_cancellation_orders,in_training_update_grad_gamma_beta,DYN_OFF,"(11, 10, 10)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,3,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"PASS,PASS",PASS,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(4, 0, 24, 0, 1, 0, 8, 4, 0, 8, 3, 2113929216, 1040187392, 1090519040)",,,"(16777216,)",768,192,Ascend950PR_9579,{},"{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'stat_rel_err', 'mere': 0.0, 'mare': 0.0, 'threshold': 0.0001, 'pass': True}, 1: {'standard': 'stat_rel_err', 'mere': 0.0, 'mare': 0.0, 'threshold': 0.0001, 'pass': True}}}",UNKNOWN,"('float32', 'float32')","((4, 24, 1, 1), (4, 24, 1, 1))","((1, 24, 1, 1), (1, 24, 1, 1))","('ND', 'ND')","('ND', 'ND')",{},{},{},{},{},{},{},{},{},{},"((4, 24, 1, 1), (4, 24, 1, 1))","('float32', 'float32')","((1, 24, 1, 1), (1, 24, 1, 1))","('ND', 'ND')","('ND', 'ND')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,all 24 permutations of extreme cancellation,,0,,(),(),()
3+in_training_update_grad_gamma_beta_ndhwc_scale_boundary_neighbors,in_training_update_grad_gamma_beta,DYN_OFF,"(12, 11, 12)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,3,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"PASS,PASS",PASS,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(8193, 0, 17, 0, 1, 0, 8, 4095, 0, 8, 3, 2013265920, 939524096, 1191182336)",,,"(16777216,)",1114248,136,Ascend950PR_9579,{},"{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'stat_rel_err', 'mere': 1.4023767244979832e-08, 'mare': 5.959009818814892e-08, 'threshold': 0.0001, 'pass': True}, 1: {'standard': 'stat_rel_err', 'mere': 1.4023767244979832e-08, 'mare': 5.959009818814892e-08, 'threshold': 0.0001, 'pass': True}}}",UNKNOWN,"('float32', 'float32')","((8193, 1, 1, 1, 17), (8193, 1, 1, 1, 17))","((1, 1, 1, 1, 17), (1, 1, 1, 1, 17))","('NDHWC', 'NDHWC')","('NDHWC', 'NDHWC')",{},{},{},{},{},{},{},{},{},{},"((8193, 1, 1, 1, 17), (8193, 1, 1, 1, 17))","('float32', 'float32')","((1, 1, 1, 1, 17), (1, 1, 1, 1, 17))","('NDHWC', 'NDHWC')","('NDHWC', 'NDHWC')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,power-of-two row-count neighbor with output tail,,0,,(),(),()
4+in_training_update_grad_gamma_beta_ncdhw_scale_boundary_neighbors,in_training_update_grad_gamma_beta,DYN_OFF,"(19, 12, 30)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,2,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"PASS,PASS",PASS,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(10000, 0, 13, 0, 1, 0, 8, 4095, 0, 8, 2, 2013265920, 939524096, 1191182336)",,,"(16777216,)",1040000,104,Ascend950PR_9579,{},"{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'stat_rel_err', 'mere': 1.3754917955566372e-08, 'mare': 3.805160749266192e-08, 'threshold': 0.0001, 'pass': True}, 1: {'standard': 'stat_rel_err', 'mere': 1.3754917955566372e-08, 'mare': 3.805160749266192e-08, 'threshold': 0.0001, 'pass': True}}}",UNKNOWN,"('float32', 'float32')","((10000, 13, 1, 1, 1), (10000, 13, 1, 1, 1))","((1, 13, 1, 1, 1), (1, 13, 1, 1, 1))","('NCDHW', 'NCDHW')","('NCDHW', 'NCDHW')",{},{},{},{},{},{},{},{},{},{},"((10000, 13, 1, 1, 1), (10000, 13, 1, 1, 1))","('float32', 'float32')","((1, 13, 1, 1, 1), (1, 13, 1, 1, 1))","('NCDHW', 'NCDHW')","('NCDHW', 'NCDHW')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,scaled and unscaled columns with one ULP neighbors,,0,,(),(),()
5+in_training_update_grad_gamma_beta_nhwc_overflow_boundary_below,in_training_update_grad_gamma_beta,DYN_OFF,"(11, 10, 10)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,2,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"PASS,PASS",PASS,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(10000, 0, 13, 0, 1, 0, 8, 4095, 0, 8, 2, 2013265920, 939524096, 1191182336)",,,"(16777216,)",1040000,104,Ascend950PR_9579,{},"{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'stat_rel_err', 'mere': 8.032658577886195e-09, 'mare': 8.032658577886193e-09, 'threshold': 0.0001, 'pass': True}, 1: {'standard': 'stat_rel_err', 'mere': 8.032658577886195e-09, 'mare': 8.032658577886193e-09, 'threshold': 0.0001, 'pass': True}}}",UNKNOWN,"('float32', 'float32')","((10000, 1, 1, 13), (10000, 1, 1, 13))","((1, 1, 1, 13), (1, 1, 1, 13))","('NHWC', 'NHWC')","('NHWC', 'NHWC')",{},{},{},{},{},{},{},{},{},{},"((10000, 1, 1, 13), (10000, 1, 1, 13))","('float32', 'float32')","((1, 1, 1, 13), (1, 1, 1, 13))","('NHWC', 'NHWC')","('NHWC', 'NHWC')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,one ULP below maximum-sum boundary,,0,,(),(),()
6+in_training_update_grad_gamma_beta_nchw_overflow_boundary_equal,in_training_update_grad_gamma_beta,DYN_OFF,"(11, 9, 9)",DYN_OFF,CST_OFF,INTrainingUpdateGradGammaBeta_cb98ad5dbefff9408f4c17988b7313f0_high_performance,DYN_OFF,CST_OFF,2,DYN_OFF,CST_OFF,UNKNOWN,DYN_OFF,CST_OFF,BINARY_MATCH,FAIL,DYN_OFF,CST_OFF,"PASS,PASS",PASS,UNKNOWN,UNKNOWN,"OK,OK,OK",PASS,DYN_OFF (DYN_OFF),"('DYN_OFF', False)",0 (0x0000000000000000),"(10000, 0, 13, 0, 1, 0, 8, 4095, 0, 8, 2, 2013265920, 939524096, 1191182336)",,,"(16777216,)",1040000,104,Ascend950PR_9579,{},"{'dyn': {}, 'cst': {}, 'bin': {0: {'standard': 'stat_rel_err', 'mere': 2.1187590649345816e-08, 'mare': 2.1187590649345813e-08, 'threshold': 0.0001, 'pass': True}, 1: {'standard': 'stat_rel_err', 'mere': 2.1187590649345816e-08, 'mare': 2.1187590649345813e-08, 'threshold': 0.0001, 'pass': True}}}",UNKNOWN,"('float32', 'float32')","((10000, 13, 1, 1), (10000, 13, 1, 1))","((1, 13, 1, 1), (1, 13, 1, 1))","('NCHW', 'NCHW')","('NCHW', 'NCHW')",{},{},{},{},{},{},{},{},{},{},"((10000, 13, 1, 1), (10000, 13, 1, 1))","('float32', 'float32')","((1, 13, 1, 1), (1, 13, 1, 1))","('NCHW', 'NCHW')","('NCHW', 'NCHW')","((0, 0), (0, 0))","((0.0001, 0.0001), (0.0001, 0.0001))","(1e-08, 1e-08)",(),(),True,finite sum near FP32 maximum,,0,,(),(),()
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/.gitattributes+2-0
@@ -0,0 +1,2 @@
1+# Preserve exact source evidence and checksum manifests.
2+** -text
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/SUMMARY.md+51-0
@@ -0,0 +1,51 @@
1+# GEIR50 同次三方精度与性能补测
2+ 
3+2026-09-28 完成原50例补测：**50/50例、100/100输出 cross_check PASS，FAIL=0**。
4+每例同一次GEIR调用均取得Torch GPU profiler `device_us`和NPU profiler原始SQLite。
5+GPU/NPU最小 **0.842386**，50例均严格大于0.1。原调用通路检查脚本实际退出0，
6+连通、三方精度、三方性能均报告50/50；本轮没有发生旧schema误判。
7+ 
8+最小比值用例为 `in_training_update_grad_gamma_beta_gen0142_nchw_f32`：
9+GPU 25.9 us / NPU 30.746 us。不要将旧GEIR `stat_rel_err`成绩混入本轮。
10+ 
11+## 证据入口
12+ 
13+- `verification_summary.json`：严格汇总结论、实际检查器版本/退出码、限制。
14+- `geir50_cases.csv`：原始50例输入，SHA256 `b9b6daa1d803106d703e23ee6b0720423159ab11dfa82b9994194816ca0f2b2c`。
15+- `geir50_result.csv`：五个已接受批次的派生合并结果；原始批次CSV保存在`runs/`。
16+- `geir50_crosscheck_outputs.csv`：100条逐输出记录，明确列出MARE/MERE/RMSE **比值**、阈值、小值域误差计数及判定。
17+- `geir50_audit.csv`：50条GPU/NPU计时、比值、原始及归档SQLite路径和SHA256。
18+- `chunk_source_map.csv`：规范用例名与原始`auto_testcase_name_N`逐条映射、原始结果SHA。
19+- `runtime_provenance.json`、`runtime_libraries_*.json`：50例对应GEIR进程实际加载私有`custom_nn`的proto/master库。
20+- `freshness.json`：Golden、源码HEAD、5个kernel部署文件哈希一致性、私有库哈希与注册隔离状态。
21+- `check_invoke_path.log`：原脚本实际输出；脚本SHA256为`a897c904fd71dde054f4ae142c1618f4bde53187128a5ed3790003d9f11252b6`。
22+- `sha256_manifest.json`：本目录文件清单和校验值（清单自身与锁文件除外）。
23+ 
24+## 方法与审计边界
25+ 
26+五批各10例串行，经强制技能入口`/workspace/ascendc-ttk-devops/scripts/run_ttk_test.sh geir`，
27+显式`--pc=1 --compare cross_check --xpu-perf --task-prof=true`，固定仓内Golden插件。
28+所有构建并行环境上限8；未运行Kernel全量、ACLNN或Torch算子接口测试。
29+每批严格核对输入/输出shape、dtype、format、origin元数据、attrs及输入范围后才绑定源用例名。
30+Golden SHA256始终为`69afaa6f90220dbf26e2d3fd964a8489915a9ea2921bfe13a84a02c3e986559d`。
31+TTK HEAD为`9bfa978d8098e2c96fd33380abab0a3c9dfd120c`，算子仓HEAD为`41169c73072287e8def27bf3bc606fa16869b63b`。
32+ 
33+50条原始TTK `cst_perf_us`均为`CST_OFF`，不能写成TTK原生性能PASS。
34+NPU时间来自**本轮**`ai_core_op_summary.db`中`task_time.duration_time`，ns除以1000转us；
35+每例1个kernel task。GPU时间来自同次结果的Torch `device_us`且status=PASS。
36+ 
37+首批attempt_01/02分别遭遇外来GPU进程，被中止并标记invalid，不纳入结论。
38+之后采用共享性能锁、每15秒资源门禁、连续至少120秒空闲才开批。
39+首批attempt_03初审发现msprof export重新生成旧中断会话的DB，导致旧DB的mtime变新；
40+已按PROF会话时间、本轮GEIR PID/用例映射及DBmtime联合筛选，离线复核接受，未重跑该批。
41+原始拒绝日志及说明在`runs/chunk_00/attempt_03/audit_resolution.md`。
42+ 
43+GPU始终为`gpu-20102`的NVIDIA A100-PCIE-40GB，UUID `GPU-0d18a557-af5c-acc4-b92a-4758753f2b82`，
44+本轮实际驱动595.84、PyTorch 2.13.0+cu130、CUDA runtime 13.0，PyTorch git为
45+`cf30153c4c131c8164ee7798e5022d810682e2cb`。前后环境快照均保存。
46+另一台20101只做只读核对，存在未映射占用且软件版本不同，未迁移服务或混用成绩。
47+ 
48+私有CANN的built-in同名注册仍采用既有字符串补丁，屏蔽范围包含REG_OP。
49+本轮证明该私有环境中的部署源码一致性、vendor加载、GEIR执行和三方精度/计时；
50+**不代表正式canndev源码构建通过，也不宣称vendor库由Golden-only提交41169c730完整重建**。
51+没有修改生产源码、Golden、他人进程或公共CANN；未提交或push。
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/capture_runtime.py+34-0
@@ -0,0 +1,34 @@
1+"""Capture only our GEIR process mappings without altering any process."""
2+import json
3+import time
4+from pathlib import Path
5+ 
6+root = Path(__file__).resolve().parent
7+project = root.parents[1]
8+seen = set()
9+while True:
10+ if (root / 'driver.pid').exists():
11+ driver = (root / 'driver.pid').read_text().strip()
12+ if driver and Path('/proc/' + driver).exists():
13+ break
14+ time.sleep(1)
15+while Path('/proc/' + driver).exists():
16+ for p in Path('/proc').iterdir():
17+ if not p.name.isdigit() or p.name in seen:
18+ continue
19+ try:
20+ cmd = (p / 'cmdline').read_bytes().replace(b'\x00', b' ').decode(errors='replace')
21+ if not cmd.startswith(str(project / 'ops-test-kit/geir/ops/in_training_update_grad_gamma_beta')):
22+ continue
23+ maps = (p / 'maps').read_text()
24+ if 'libcust_opsproto_rt2.0.so' not in maps:
25+ continue
26+ libs = sorted({line.split()[-1] for line in maps.splitlines()
27+ if '/Ascend/cann-9.2.0/' in line})
28+ (root / ('runtime_libraries_' + p.name + '.json')).write_text(json.dumps({
29+ 'pid': int(p.name), 'cmdline': cmd, 'libraries': libs,
30+ 'captured_ns': time.time_ns()}, indent=2) + '\n')
31+ seen.add(p.name)
32+ except (OSError, UnicodeError):
33+ pass
34+ time.sleep(1)
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/check_invoke_path.log+6-0
@@ -0,0 +1,6 @@
1+用例数 : 50 (门槛 50)
2+通路连通 : 50/50
3+三方精度跑到 : 50/50
4+三方性能跑到 : 50/50
5+ 
6+✓ 达标：通路连通 + 三方精度 + 三方性能，三项全覆盖
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/check_invoke_path_used.py+220-0
@@ -0,0 +1,220 @@
1+#!/usr/bin/env python3
2+"""通路验证门禁：从 TTK 结果 CSV 证明【三方精度】和【三方性能】真的跑到了。
3+ 
4+为什么需要这个：两项都会【静默失效】——
5+ · third_party 没挂 / provider 解析失败 → precision_metrics 里是
6+ {'standard':'cross_check','pass':False,'reason':'third_party unavailable'}，
7+ 看总体 PASS/FAIL 根本发现不了少了一条腿；
8+ · --xpu-perf 没传（或该通路 CLI 未注册该参数）→ xpu_metrics 里只有
9+ {'status':'PASS','api':...}【没有 device_us】，一样看不出来。
10+ 
11+判据（均取自实测的结果 CSV 字段，不是推断）：
12+ 三方精度跑到 = precision_metrics 中该输出 standard=='cross_check' 且带 'result'
13+ （含 mare/mere/rmse 实测值）
14+ 三方性能跑到 = xpu_metrics 的 provider 字典里有 'device_us' 且 > 0
15+ 
16+用法：
17+ python3 check_invoke_path.py <result.csv> [more.csv ...] [--min-cases 50]
18+ python3 check_invoke_path.py <result.csv> --pathway aclnn # 该通路 TTK 暂不支持三方，只查连通
19+ 
20+退出码：0=达标，1=不达标（有明细输出，直接贴进 summary 的「结论」项）
21+"""
22+import argparse
23+import ast
24+import csv
25+import sys
26+from collections import Counter
27+ 
28+# precision_metrics 的顶层按【执行段】分组，段名随通路变：
29+# kernel/geir → bin(部署 binary) / dyn(JIT 源码) / cst(常量折叠)
30+# aclnn/e2e → eager 等
31+# 故不写死段名，遍历所有段。
32+# 各通路在 invoke_path 档要验什么（handover-deliverables.md §5.1 ②）：
33+# kernel —— 【不进本档】三项各出原交付件：泛化 generalization/(~1000)、
34+# 三方精度 precision/(~100)、三方性能 performance/(~200)；
35+# 连通性由泛化档那 ~1000 例天然证明，无需另交。
36+# geir —— 连通 + 三方精度 + 三方性能，一把呈现（它没有独立的三方/性能档）。
37+# aclnn/e2e —— 只验连通：TTK 尚未在这两条通路派发三方
38+# （framework_api 硬编码 third_parties=None；aclnn 无 dispatch_xpu 调用点）。
39+_NO_3PARTY_PATHWAYS = {"aclnn", "e2e"}
40+_SKIP_REASON = {
41+ "aclnn": "TTK 当前不在该通路派发三方，三方两列填 N/A(TTK暂不支持)，只验通路连通",
42+ "e2e": "TTK 当前不在该通路派发三方，三方两列填 N/A(TTK暂不支持)，只验通路连通",
43+}
44+ 
45+ 
46+def _lit(s):
47+ try:
48+ return ast.literal_eval(s) if s and s.strip() not in ("", "NA") else None
49+ except (ValueError, SyntaxError):
50+ return None
51+ 
52+ 
53+def _scan_precision(cell):
54+ """→ (三方精度是否跑到, 整数硬路由输出数, 失败原因)"""
55+ m = _lit(cell)
56+ if not isinstance(m, dict):
57+ return False, 0, "precision_metrics 不可解析" if cell else "precision_metrics 为空"
58+ hit, int_routed, reason = False, 0, ""
59+ for _seg, outs in m.items():
60+ if not isinstance(outs, dict):
61+ continue
62+ for _idx, d in outs.items():
63+ if not isinstance(d, dict):
64+ continue
65+ std = d.get("standard")
66+ if std == "cross_check":
67+ if isinstance(d.get("result"), dict):
68+ hit = True
69+ else:
70+ reason = reason or d.get("reason") or "cross_check 无 result"
71+ elif std in ("binary_equal", "requant"):
72+ # 不参与三方的输出，不算缺腿：
73+ # binary_equal —— 整数/bool 输出（精确整数语义，逐位相等）
74+ # requant —— 浮点输入 + int4/int8 输出的量化场景，标准是单标杆
75+ # 绝对误差 <= 1，明确不做三方
76+ int_routed += 1
77+ if not hit and not reason:
78+ reason = "无任何输出走 cross_check（tolerance 没配 / 全是整数或量化输出）"
79+ return hit, int_routed, reason
80+ 
81+ 
82+def _scan_xpu(cell):
83+ """→ (三方性能是否跑到, device_us, 失败原因)"""
84+ m = _lit(cell)
85+ if not isinstance(m, dict):
86+ txt = (cell or "").strip()
87+ return False, None, f"xpu_metrics={txt[:40]}" if txt else "xpu_metrics 为空（未连远端）"
88+ for _prov, d in m.items():
89+ if not isinstance(d, dict):
90+ continue
91+ us = d.get("device_us")
92+ if isinstance(us, (int, float)) and us > 0:
93+ return True, us, ""
94+ return False, None, "有 provider 但无 device_us（--xpu-perf 未生效）"
95+ return False, None, "xpu_metrics 为空字典"
96+ 
97+ 
98+_BAD_STATUS = ("EXEC_FAILURE", "UNSUPPORTED", "GOLDEN_FAILURE", "SOC_NOT_SUPPORT",
99+ "COMPILE_FAILURE", "PROFILE_CRASH", "ERROR")
100+ 
101+ 
102+def _scan_geir(row):
103+ """geir 结果表【没有 precision_metrics 列】，只有 precision/precision_status +
104+ xpu_metrics。故按该 schema 判：
105+ 
106+ 连通 = precision_status 不是 EXEC_FAILURE/UNSUPPORTED/... 这类硬失败
107+ 三方性能 = xpu_metrics 里有 device_us>0
108+ 三方精度 = 三方数据已到位（xpu_metrics 有值，与 cross_check 同一次 collect_third_party
109+ 取回）且精度判定不是 GOLDEN_FAILURE —— cross_check 拿不到三方时会直接
110+ 判失败，故 PASS/FAIL 出得来即证明三方腿在。
111+ ⚠️ 这是间接判据：geir 不输出逐输出明细，无法像 kernel 那样直接看到
112+ standard=='cross_check' 的 result。
113+ """
114+ st = (row.get("precision_status") or "").strip().upper()
115+ prec = (row.get("precision") or "").strip().upper()
116+ hard_fail = any(b in st or b in prec for b in _BAD_STATUS)
117+ connected = bool(st) and not hard_fail
118+ x_hit, us, x_why = _scan_xpu(row.get("xpu_metrics", ""))
119+ p_hit = connected and x_hit
120+ p_why = "" if p_hit else ("三方数据未到位（xpu_metrics 空）" if not x_hit
121+ else f"精度判定硬失败 precision_status={st}")
122+ return connected, p_hit, p_why, x_hit, x_why
123+ 
124+ 
125+def main():
126+ ap = argparse.ArgumentParser()
127+ ap.add_argument("csvs", nargs="+")
128+ ap.add_argument("--min-cases", type=int, default=50)
129+ ap.add_argument("--pathway", default="", required=True,
130+ choices=["geir", "aclnn", "e2e"],
131+ help="必填：决定该通路要验哪几项（见 handover-deliverables.md §5.1 ②）。"
132+ "kernel 不进 invoke_path 档，故不在选项内")
133+ args = ap.parse_args()
134+ 
135+ skip_3p = args.pathway in _NO_3PARTY_PATHWAYS
136+ total = conn_ok = prec_ok = perf_ok = int_routed = prec_exempt = 0
137+ reasons = Counter()
138+ 
139+ for path in args.csvs:
140+ with open(path, newline="", encoding="utf-8-sig") as f:
141+ for row in csv.DictReader(f):
142+ total += 1
143+ if "precision_metrics" in row:
144+ # kernel/aclnn schema：有逐输出明细，可直接看 standard/result
145+ # 连通 = 真的算出了判定结果；precision_metrics 为空 {} 说明
146+ # golden 没生成 / 接口没执行成功，此时哪怕三方免检也不能算通过。
147+ pm = _lit(row.get("precision_metrics", ""))
148+ connected = bool(pm) and any(
149+ isinstance(v, dict) and v for v in pm.values())
150+ conn_why = (f"precision_metrics 为空（golden 未生成/接口未执行成功，"
151+ f"precision_status={row.get('precision_status', '?')}）")
152+ p_hit, n_int, p_why = _scan_precision(row.get("precision_metrics", ""))
153+ x_hit, _us, x_why = _scan_xpu(row.get("xpu_metrics", ""))
154+ else:
155+ # geir schema：无 precision_metrics 列，按 precision_status + xpu_metrics 判
156+ connected, p_hit, p_why, x_hit, x_why = _scan_geir(row)
157+ n_int = 0
158+ conn_why = (f"precision={row.get('precision', '?')[:60]} "
159+ f"(status={row.get('precision_status', '?')[:40]})")
160+ conn_ok += connected
161+ prec_ok += p_hit
162+ perf_ok += x_hit
163+ int_routed += n_int
164+ # 该用例【所有输出】都被标准硬路由到 binary_equal/requant(整数或量化输出)时,
165+ # 不走 cross_check 是判据本身的规定, 不是"三方腿断了" —— 记豁免, 不记失败。
166+ # (区分依据: n_int>0 说明确实解析到了逐输出判定, 只是没有一个该走三方)
167+ exempt = (not p_hit) and n_int > 0
168+ prec_exempt += exempt
169+ if not connected:
170+ reasons[f"[通路连通] {conn_why}"] += 1
171+ elif not skip_3p:
172+ if not p_hit and not exempt:
173+ reasons[f"[三方精度] {p_why}"] += 1
174+ if not x_hit:
175+ reasons[f"[三方性能] {x_why}"] += 1
176+ 
177+ if total == 0:
178+ print("✗ 结果 CSV 为空——通路根本没跑通（先查 op_name 是否写成驼峰）")
179+ return 1
180+ 
181+ print(f"用例数 : {total} (门槛 {args.min_cases})")
182+ print(f"通路连通 : {conn_ok}/{total}")
183+ print(f"三方精度跑到 : {prec_ok}/{total}")
184+ print(f"三方性能跑到 : {perf_ok}/{total}")
185+ if int_routed:
186+ print(f" └ 另有 {int_routed} 个输出不参与三方（整数 binary_equal / 量化 requant，正常）")
187+ if prec_exempt:
188+ print(f" └ 其中 {prec_exempt}/{total} 例的输出【全部】按标准硬路由到 binary_equal/requant，"
189+ f"三方精度对这些用例不适用（豁免，不算缺腿）")
190+ if reasons:
191+ print("未跑到的原因分布：")
192+ for why, n in reasons.most_common(8):
193+ print(f" {n:5d} {why}")
194+ 
195+ fail = []
196+ if total < args.min_cases:
197+ fail.append(f"用例数 {total} < {args.min_cases}")
198+ if conn_ok < total:
199+ fail.append(f"通路连通只有 {conn_ok}/{total}")
200+ if skip_3p:
201+ print(f"\n注：--pathway={args.pathway} —— {_SKIP_REASON[args.pathway]}。")
202+ else:
203+ if prec_ok + prec_exempt < total:
204+ fail.append(f"三方精度只跑到 {prec_ok}/{total}（豁免 {prec_exempt}）")
205+ if perf_ok < total:
206+ fail.append(f"三方性能只跑到 {perf_ok}/{total}")
207+ 
208+ if fail:
209+ print("\n✗ 不达标：" + "；".join(fail))
210+ return 1
211+ if skip_3p:
212+ print(f"\n✓ 达标：通路连通 {conn_ok}/{total}，用例数 {total} ≥ {args.min_cases}")
213+ else:
214+ tail = f"（其中 {prec_exempt} 例三方精度按标准豁免）" if prec_exempt else ""
215+ print(f"\n✓ 达标：通路连通 + 三方精度 + 三方性能，三项全覆盖{tail}")
216+ return 0
217+ 
218+ 
219+if __name__ == "__main__":
220+ sys.exit(main())
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunk_source_map.csv+51-0
@@ -0,0 +1,51 @@
1+testcase_name,original_testcase_name,chunk,attempt,original_result,original_result_sha256,original_audit,original_audit_sha256
2+in_training_update_grad_gamma_beta_gen0000_nchw_f32,auto_testcase_name_1,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
3+in_training_update_grad_gamma_beta_gen0020_nchw_f32,auto_testcase_name_2,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
4+in_training_update_grad_gamma_beta_gen0041_nhwc_f32,auto_testcase_name_3,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
5+in_training_update_grad_gamma_beta_gen0061_nhwc_f32,auto_testcase_name_4,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
6+in_training_update_grad_gamma_beta_gen0081_nhwc_f32,auto_testcase_name_5,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
7+in_training_update_grad_gamma_beta_gen0102_nchw_f32,auto_testcase_name_6,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
8+in_training_update_grad_gamma_beta_gen0122_nchw_f32,auto_testcase_name_7,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
9+in_training_update_grad_gamma_beta_gen0142_nchw_f32,auto_testcase_name_8,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
10+in_training_update_grad_gamma_beta_gen0162_nchw_f32,auto_testcase_name_9,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
11+in_training_update_grad_gamma_beta_gen0184_nchw_f32,auto_testcase_name_10,chunk_00,attempt_03,runs/chunk_00/attempt_03/result.csv,83c15537c69027cd1f4c07ac915b2d45613cdc2fa16a2c454c61b1feb304d0ff,runs/chunk_00/attempt_03/audit.csv,598b646663538b77c1106592674b3b38bbed63e0ab1936f006294938b87cd3a0
12+in_training_update_grad_gamma_beta_gen0204_nchw_f32,auto_testcase_name_1,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
13+in_training_update_grad_gamma_beta_gen0224_nchw_f32,auto_testcase_name_2,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
14+in_training_update_grad_gamma_beta_gen0245_nhwc_f32,auto_testcase_name_3,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
15+in_training_update_grad_gamma_beta_gen0265_nhwc_f32,auto_testcase_name_4,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
16+in_training_update_grad_gamma_beta_gen0285_nhwc_f32,auto_testcase_name_5,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
17+in_training_update_grad_gamma_beta_gen0306_nchw_f32,auto_testcase_name_6,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
18+in_training_update_grad_gamma_beta_gen0326_nchw_f32,auto_testcase_name_7,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
19+in_training_update_grad_gamma_beta_gen0346_nchw_f32,auto_testcase_name_8,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
20+in_training_update_grad_gamma_beta_gen0367_nhwc_f32,auto_testcase_name_9,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
21+in_training_update_grad_gamma_beta_gen0387_nhwc_f32,auto_testcase_name_10,chunk_01,attempt_01,runs/chunk_01/attempt_01/result.csv,aebd153d334c03210b9f5fbc645c01da5b4b10b990987c6fdcef13e0b5e8a39d,runs/chunk_01/attempt_01/audit.csv,c2889cb9d92af67e2aa35f534e7704fd80c9f3640285d6daaa0e0133e62e94a0
22+in_training_update_grad_gamma_beta_gen0407_nhwc_f32,auto_testcase_name_1,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
23+in_training_update_grad_gamma_beta_gen0427_nhwc_f32,auto_testcase_name_2,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
24+in_training_update_grad_gamma_beta_gen0449_nhwc_f32,auto_testcase_name_3,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
25+in_training_update_grad_gamma_beta_gen0469_nhwc_f32,auto_testcase_name_4,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
26+in_training_update_grad_gamma_beta_gen0489_nhwc_f32,auto_testcase_name_5,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
27+in_training_update_grad_gamma_beta_gen0510_nchw_f32,auto_testcase_name_6,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
28+in_training_update_grad_gamma_beta_gen0530_nchw_f32,auto_testcase_name_7,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
29+in_training_update_grad_gamma_beta_gen0550_nchw_f32,auto_testcase_name_8,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
30+in_training_update_grad_gamma_beta_gen0571_nhwc_f32,auto_testcase_name_9,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
31+in_training_update_grad_gamma_beta_gen0591_nhwc_f32,auto_testcase_name_10,chunk_02,attempt_01,runs/chunk_02/attempt_01/result.csv,5b927b7a578e0be4fdab7eca0e2e5588a42487add11a95de900914b93c62b7d4,runs/chunk_02/attempt_01/audit.csv,e1f6830aef47e92c4ed25ed948dd9cb60fa9f8dedc630d060a50860af5f36138
32+in_training_update_grad_gamma_beta_gen0611_nhwc_f32,auto_testcase_name_1,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
33+in_training_update_grad_gamma_beta_gen0631_nhwc_f32,auto_testcase_name_2,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
34+in_training_update_grad_gamma_beta_gen0652_nchw_f32,auto_testcase_name_3,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
35+in_training_update_grad_gamma_beta_gen0672_nchw_f32,auto_testcase_name_4,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
36+in_training_update_grad_gamma_beta_gen0692_nchw_f32,auto_testcase_name_5,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
37+in_training_update_grad_gamma_beta_gen0714_nchw_f32,auto_testcase_name_6,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
38+in_training_update_grad_gamma_beta_gen0734_nchw_f32,auto_testcase_name_7,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
39+in_training_update_grad_gamma_beta_gen0754_nchw_f32,auto_testcase_name_8,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
40+in_training_update_grad_gamma_beta_gen0775_nhwc_f32,auto_testcase_name_9,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
41+in_training_update_grad_gamma_beta_gen0795_nhwc_f32,auto_testcase_name_10,chunk_03,attempt_01,runs/chunk_03/attempt_01/result.csv,bf14cf473d73535aa9b6db927b671d3abf0470095edc8847e358e6da1e8df555,runs/chunk_03/attempt_01/audit.csv,8a97431c635f4048e5cefecbcca73a42e66eb29e8b8c3beb0ac03de39459b0c7
42+in_training_update_grad_gamma_beta_gen0815_nhwc_f32,auto_testcase_name_1,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
43+in_training_update_grad_gamma_beta_gen0836_nchw_f32,auto_testcase_name_2,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
44+in_training_update_grad_gamma_beta_gen0856_nchw_f32,auto_testcase_name_3,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
45+in_training_update_grad_gamma_beta_gen0876_nchw_f32,auto_testcase_name_4,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
46+in_training_update_grad_gamma_beta_gen0896_nchw_f32,auto_testcase_name_5,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
47+in_training_update_grad_gamma_beta_gen0917_nhwc_f32,auto_testcase_name_6,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
48+in_training_update_grad_gamma_beta_gen0937_nhwc_f32,auto_testcase_name_7,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
49+in_training_update_grad_gamma_beta_gen0957_nhwc_f32,auto_testcase_name_8,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
50+in_training_update_grad_gamma_beta_gen0979_nhwc_f32,auto_testcase_name_9,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
51+in_training_update_grad_gamma_beta_gen0999_nhwc_f32,auto_testcase_name_10,chunk_04,attempt_01,runs/chunk_04/attempt_01/result.csv,41854815fa540c2c0dcb21ac5d4cd147ee10258916f60736364bda0c72b083cc,runs/chunk_04/attempt_01/audit.csv,db004faba919d62de8337151d4d5f1d81ed174f0837a47c63d638cc8af5d4ddc
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_00.csv+11-0
@@ -0,0 +1,11 @@
1+testcase_name,network_name,op_name,input_shapes,input_dtypes,input_formats,output_shapes,output_dtypes,output_formats,input_ori_shapes,input_ori_formats,output_ori_shapes,output_ori_formats,attributes,input_data_ranges,precision_tolerances,absolute_precision,output_inplace_indexes,output_shape_unknown_indexes,is_enabled,remark,soc_series,priority,dump_file_prefix,manual_input_binaries,manual_golden_binaries
2+in_training_update_grad_gamma_beta_gen0000_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[1, 1, 1, 1], [1, 1, 1, 1]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 1, 1, 1], [1, 1, 1, 1]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 1, 1, 1], [1, 1, 1, 1]]","['NCHW', 'NCHW']","[[1, 1, 1, 1], [1, 1, 1, 1]]","['NCHW', 'NCHW']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
3+in_training_update_grad_gamma_beta_gen0020_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[2, 1, 1, 648], [2, 1, 1, 648]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 1, 1, 648], [1, 1, 1, 648]]","['float32', 'float32']","['NCHW', 'NCHW']","[[2, 1, 1, 648], [2, 1, 1, 648]]","['NCHW', 'NCHW']","[[1, 1, 1, 648], [1, 1, 1, 648]]","['NCHW', 'NCHW']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
4+in_training_update_grad_gamma_beta_gen0041_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[127, 5, 15, 1], [127, 5, 15, 1]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 5, 15, 1], [1, 5, 15, 1]]","['float32', 'float32']","['NHWC', 'NHWC']","[[127, 5, 15, 1], [127, 5, 15, 1]]","['NHWC', 'NHWC']","[[1, 5, 15, 1], [1, 5, 15, 1]]","['NHWC', 'NHWC']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
5+in_training_update_grad_gamma_beta_gen0061_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[16, 7, 5, 7], [16, 7, 5, 7]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 7, 5, 7], [1, 7, 5, 7]]","['float32', 'float32']","['NHWC', 'NHWC']","[[16, 7, 5, 7], [16, 7, 5, 7]]","['NHWC', 'NHWC']","[[1, 7, 5, 7], [1, 7, 5, 7]]","['NHWC', 'NHWC']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
6+in_training_update_grad_gamma_beta_gen0081_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[31, 7, 3, 127], [31, 7, 3, 127]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 7, 3, 127], [1, 7, 3, 127]]","['float32', 'float32']","['NHWC', 'NHWC']","[[31, 7, 3, 127], [31, 7, 3, 127]]","['NHWC', 'NHWC']","[[1, 7, 3, 127], [1, 7, 3, 127]]","['NHWC', 'NHWC']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
7+in_training_update_grad_gamma_beta_gen0102_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[8, 5, 3, 65], [8, 5, 3, 65]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 5, 3, 65], [1, 5, 3, 65]]","['float32', 'float32']","['NCHW', 'NCHW']","[[8, 5, 3, 65], [8, 5, 3, 65]]","['NCHW', 'NCHW']","[[1, 5, 3, 65], [1, 5, 3, 65]]","['NCHW', 'NCHW']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
8+in_training_update_grad_gamma_beta_gen0122_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[5, 3, 32, 16], [5, 3, 32, 16]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 3, 32, 16], [1, 3, 32, 16]]","['float32', 'float32']","['NCHW', 'NCHW']","[[5, 3, 32, 16], [5, 3, 32, 16]]","['NCHW', 'NCHW']","[[1, 3, 32, 16], [1, 3, 32, 16]]","['NCHW', 'NCHW']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
9+in_training_update_grad_gamma_beta_gen0142_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[16, 7, 32, 648], [16, 7, 32, 648]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 7, 32, 648], [1, 7, 32, 648]]","['float32', 'float32']","['NCHW', 'NCHW']","[[16, 7, 32, 648], [16, 7, 32, 648]]","['NCHW', 'NCHW']","[[1, 7, 32, 648], [1, 7, 32, 648]]","['NCHW', 'NCHW']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
10+in_training_update_grad_gamma_beta_gen0162_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[7, 7, 32, 4], [7, 7, 32, 4]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 7, 32, 4], [1, 7, 32, 4]]","['float32', 'float32']","['NCHW', 'NCHW']","[[7, 7, 32, 4], [7, 7, 32, 4]]","['NCHW', 'NCHW']","[[1, 7, 32, 4], [1, 7, 32, 4]]","['NCHW', 'NCHW']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
11+in_training_update_grad_gamma_beta_gen0184_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[127, 7, 1, 5], [127, 7, 1, 5]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 7, 1, 5], [1, 7, 1, 5]]","['float32', 'float32']","['NCHW', 'NCHW']","[[127, 7, 1, 5], [127, 7, 1, 5]]","['NCHW', 'NCHW']","[[1, 7, 1, 5], [1, 7, 1, 5]]","['NCHW', 'NCHW']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_00_run.log+264-0
@@ -0,0 +1,264 @@
1+2026-09-28 01:31:23,248 [DEBUG] [1673918 MainProcess MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
2+2026-09-28 01:31:23,249 [INFO] [1673918 MP MT] [bridge.py:441]: Command: ttk geir -i /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_00.csv
3+2026-09-28 01:31:23,251 [INFO] [1673918 MP MT] [instance_base.py:118]: TTK Profiling Start!
4+2026-09-28 01:31:23,251 [INFO] [1673918 MP MT] [instance_base.py:119]: Mode: ASCEND_ONBOARD
5+2026-09-28 01:31:23,254 [INFO] [1673918 MP MT] [instance_base.py:122]: Device Platform: Ascend950PR_9579
6+2026-09-28 01:31:23,255 [INFO] [1673918 MP MT] [instance_base.py:129]: Device Count: 1
7+2026-09-28 01:31:23,255 [INFO] [1673918 MP MT] [instance_base.py:372]: Preparing testcases ...
8+2026-09-28 01:31:23,255 [INFO] [1673918 MP MT] [instance_base.py:374]: Reading testcase file /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_00.csv ...
9+2026-09-28 01:31:23,255 [INFO] [1673918 MP MT] [instance_base.py:447]: Input testcase file is not a valid zip file, switch to normal csv mode...
10+2026-09-28 01:31:23,255 [INFO] [1673918 MP MT] [instance_base.py:723]: Reading testcases from /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_00.csv ...
11+2026-09-28 01:31:23,255 [WARNING] [1673918 TestcaseManager HeaderCheckTestcaseName] [testcase_manager.py:258]: Testcase name not found! It is important to add a testcase_name in order to identify your testcases
12+2026-09-28 01:31:23,255 [WARNING] [1673918 TestcaseManager HeaderCheckUnidentifiedHeaders] [testcase_manager.py:289]: Detected unidentified headers and will be ignored: ['\ufefftestcase_name']
13+2026-09-28 01:31:23,258 [INFO] [1673918 TestcaseManager auto_testcase_name_1] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
14+2026-09-28 01:31:23,310 [INFO] [1673918 MP MT] [instance_base.py:378]: Checking testcase name...
15+2026-09-28 01:31:23,310 [INFO] [1673918 MP MT] [instance_base.py:381]: Case number: 10
16+2026-09-28 01:31:23,322 [INFO] [1673918 MP MT] [heartbeat_manager.py:86]: Heartbeat subprocess started: pid=1673920, health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_1d8c46ecacc1.json
17+2026-09-28 01:31:23,324 [INFO] [1673920 HB MT] [heartbeat.py:50]: HB started: tenant=1d8c46ecacc1 health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_1d8c46ecacc1.json n_endpoints=1 tls=False
18+2026-09-28 01:31:23,360 [INFO] [1673918 MP MT] [instance_base.py:463]: Initialize output csv file...
19+2026-09-28 01:31:23,404 [INFO] [1673918 MP MT] [instance_base.py:167]: Process per device: 1
20+2026-09-28 01:31:23,404 [INFO] [1673918 MP MT] [instance_base.py:169]: Preparing Task Executors...
21+2026-09-28 01:31:23,502 [DEBUG] [1673932 D0P0 MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
22+2026-09-28 01:31:23,509 [INFO] [1673918 MP MT] [instance_base.py:301]:
23++--------------------------------------------------------------------------------------+
24+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 0s Rev: 9bfa978 |
25++--------------------------------------------------------------------------------------+
26+| NPU:0 Ascend950PR_9579 | 1673932 D0P0 UNKNOWN IDLE 0s |
27++--------------------------------------------------------------------------------------+
28+2026-09-28 01:31:23,546 [INFO] [1673932 auto_testcase_name_1 OnGenerateInput] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
29+2026-09-28 01:31:26,307 [WARNING] [1673932 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:550]: 424 retry (1/5): {"error": "missing spec dependency: golden", "missing": "golden"}
30+2026-09-28 01:31:26,388 [INFO] [1673932 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:357]: 424: synced 'golden.py' for missing module 'golden'
31+2026-09-28 01:31:30,820 [INFO] [1673932 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
32+2026-09-28 01:31:30,821 [INFO] [1673932 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_1.json
33+2026-09-28 01:31:30,821 [INFO] [1673932 auto_testcase_name_1 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
34+2026-09-28 01:31:33,508 [INFO] [1673918 MP MT] [instance_base.py:301]:
35++--------------------------------------------------------------------------------------+
36+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 10s Rev: 9bfa978 |
37++--------------------------------------------------------------------------------------+
38+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_1 OnGeirExecute RUNNING 2s |
39++--------------------------------------------------------------------------------------+
40+2026-09-28 01:31:43,513 [INFO] [1673918 MP MT] [instance_base.py:301]:
41++---------------------------------------------------------------------------------------+
42+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 20s Rev: 9bfa978 |
43++---------------------------------------------------------------------------------------+
44+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_1 OnGeirExecute RUNNING 12s |
45++---------------------------------------------------------------------------------------+
46+2026-09-28 01:31:46,649 [INFO] [1673932 auto_testcase_name_1 OnReturning] [profiling.py:63]:
47+########################
48+Performance result: Comparison result:
49+CST_PERF: None CST_GOLD: PASS,PASS
50+DYN_PERF: None DYN_GOLD:
51+PRECISION_STATUS: PASS PASSED: PASS
52+########################
53+
54+2026-09-28 01:31:51,164 [INFO] [1673932 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
55+2026-09-28 01:31:51,165 [INFO] [1673932 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_2.json
56+2026-09-28 01:31:51,165 [INFO] [1673932 auto_testcase_name_2 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
57+2026-09-28 01:31:53,513 [INFO] [1673918 MP MT] [instance_base.py:301]:
58++--------------------------------------------------------------------------------------------------------+
59+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 30s Rev: 9bfa978 |
60++--------------------------------------------------------------------------------------------------------+
61+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_2 OnGeirExecute RUNNING 2s |
62++--------------------------------------------------------------------------------------------------------+
63+2026-09-28 01:32:03,521 [INFO] [1673918 MP MT] [instance_base.py:301]:
64++---------------------------------------------------------------------------------------+
65+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 40s Rev: 9bfa978 |
66++---------------------------------------------------------------------------------------+
67+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_2 OnGeirExecute RUNNING 12s |
68++---------------------------------------------------------------------------------------+
69+2026-09-28 01:32:05,873 [INFO] [1673932 auto_testcase_name_2 OnReturning] [profiling.py:63]:
70+########################
71+Performance result: Comparison result:
72+CST_PERF: None CST_GOLD: PASS,PASS
73+DYN_PERF: None DYN_GOLD:
74+PRECISION_STATUS: PASS PASSED: PASS
75+########################
76+
77+2026-09-28 01:32:10,356 [INFO] [1673932 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
78+2026-09-28 01:32:10,356 [INFO] [1673932 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_3.json
79+2026-09-28 01:32:10,356 [INFO] [1673932 auto_testcase_name_3 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
80+2026-09-28 01:32:13,529 [INFO] [1673918 MP MT] [instance_base.py:301]:
81++--------------------------------------------------------------------------------------------------------+
82+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 50s Rev: 9bfa978 |
83++--------------------------------------------------------------------------------------------------------+
84+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_3 OnGeirExecute RUNNING 3s |
85++--------------------------------------------------------------------------------------------------------+
86+2026-09-28 01:32:23,537 [INFO] [1673918 MP MT] [instance_base.py:301]:
87++---------------------------------------------------------------------------------------+
88+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 60s Rev: 9bfa978 |
89++---------------------------------------------------------------------------------------+
90+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_3 OnGeirExecute RUNNING 13s |
91++---------------------------------------------------------------------------------------+
92+2026-09-28 01:32:26,000 [INFO] [1673932 auto_testcase_name_3 OnReturning] [profiling.py:63]:
93+########################
94+Performance result: Comparison result:
95+CST_PERF: None CST_GOLD: PASS,PASS
96+DYN_PERF: None DYN_GOLD:
97+PRECISION_STATUS: PASS PASSED: PASS
98+########################
99+
100+2026-09-28 01:32:30,471 [INFO] [1673932 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
101+2026-09-28 01:32:30,471 [INFO] [1673932 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_4.json
102+2026-09-28 01:32:30,472 [INFO] [1673932 auto_testcase_name_4 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
103+2026-09-28 01:32:33,545 [INFO] [1673918 MP MT] [instance_base.py:301]:
104++--------------------------------------------------------------------------------------------------------+
105+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 70s Rev: 9bfa978 |
106++--------------------------------------------------------------------------------------------------------+
107+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_4 OnGeirExecute RUNNING 3s |
108++--------------------------------------------------------------------------------------------------------+
109+2026-09-28 01:32:43,552 [INFO] [1673918 MP MT] [instance_base.py:301]:
110++---------------------------------------------------------------------------------------+
111+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 80s Rev: 9bfa978 |
112++---------------------------------------------------------------------------------------+
113+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_4 OnGeirExecute RUNNING 13s |
114++---------------------------------------------------------------------------------------+
115+2026-09-28 01:32:45,291 [INFO] [1673932 auto_testcase_name_4 OnReturning] [profiling.py:63]:
116+########################
117+Performance result: Comparison result:
118+CST_PERF: None CST_GOLD: PASS,PASS
119+DYN_PERF: None DYN_GOLD:
120+PRECISION_STATUS: PASS PASSED: PASS
121+########################
122+
123+2026-09-28 01:32:49,780 [INFO] [1673932 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
124+2026-09-28 01:32:49,780 [INFO] [1673932 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_5.json
125+2026-09-28 01:32:49,781 [INFO] [1673932 auto_testcase_name_5 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
126+2026-09-28 01:32:53,553 [INFO] [1673918 MP MT] [instance_base.py:301]:
127++--------------------------------------------------------------------------------------------------------+
128+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 90s Rev: 9bfa978 |
129++--------------------------------------------------------------------------------------------------------+
130+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_5 OnGeirExecute RUNNING 3s |
131++--------------------------------------------------------------------------------------------------------+
132+2026-09-28 01:33:03,559 [INFO] [1673918 MP MT] [instance_base.py:301]:
133++----------------------------------------------------------------------------------------------------------+
134+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 100s Rev: 9bfa978 |
135++----------------------------------------------------------------------------------------------------------+
136+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_5 OnGeirExecute RUNNING 13s |
137++----------------------------------------------------------------------------------------------------------+
138+2026-09-28 01:33:04,571 [INFO] [1673932 auto_testcase_name_5 OnReturning] [profiling.py:63]:
139+########################
140+Performance result: Comparison result:
141+CST_PERF: None CST_GOLD: PASS,PASS
142+DYN_PERF: None DYN_GOLD:
143+PRECISION_STATUS: PASS PASSED: PASS
144+########################
145+
146+2026-09-28 01:33:09,088 [INFO] [1673932 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
147+2026-09-28 01:33:09,088 [INFO] [1673932 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_6.json
148+2026-09-28 01:33:09,089 [INFO] [1673932 auto_testcase_name_6 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
149+2026-09-28 01:33:13,566 [INFO] [1673918 MP MT] [instance_base.py:301]:
150++---------------------------------------------------------------------------------------------------------+
151+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 110s Rev: 9bfa978 |
152++---------------------------------------------------------------------------------------------------------+
153+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_6 OnGeirExecute RUNNING 4s |
154++---------------------------------------------------------------------------------------------------------+
155+2026-09-28 01:33:23,572 [INFO] [1673918 MP MT] [instance_base.py:301]:
156++----------------------------------------------------------------------------------------------------------+
157+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 120s Rev: 9bfa978 |
158++----------------------------------------------------------------------------------------------------------+
159+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_6 OnGeirExecute RUNNING 14s |
160++----------------------------------------------------------------------------------------------------------+
161+2026-09-28 01:33:23,810 [INFO] [1673932 auto_testcase_name_6 OnReturning] [profiling.py:63]:
162+########################
163+Performance result: Comparison result:
164+CST_PERF: None CST_GOLD: PASS,PASS
165+DYN_PERF: None DYN_GOLD:
166+PRECISION_STATUS: PASS PASSED: PASS
167+########################
168+
169+2026-09-28 01:33:28,292 [INFO] [1673932 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
170+2026-09-28 01:33:28,293 [INFO] [1673932 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_7.json
171+2026-09-28 01:33:28,293 [INFO] [1673932 auto_testcase_name_7 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
172+2026-09-28 01:33:33,575 [INFO] [1673918 MP MT] [instance_base.py:301]:
173++---------------------------------------------------------------------------------------------------------+
174+| Version: 3.0.0 Summary (Device Total: 1) Progress: 60% 6 / 10 ET: 130s Rev: 9bfa978 |
175++---------------------------------------------------------------------------------------------------------+
176+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_7 OnGeirExecute RUNNING 5s |
177++---------------------------------------------------------------------------------------------------------+
178+2026-09-28 01:33:42,900 [INFO] [1673932 auto_testcase_name_7 OnReturning] [profiling.py:63]:
179+########################
180+Performance result: Comparison result:
181+CST_PERF: None CST_GOLD: PASS,PASS
182+DYN_PERF: None DYN_GOLD:
183+PRECISION_STATUS: PASS PASSED: PASS
184+########################
185+
186+2026-09-28 01:33:43,578 [INFO] [1673918 MP MT] [instance_base.py:301]:
187++---------------------------------------------------------------------------------------------------------+
188+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 140s Rev: 9bfa978 |
189++---------------------------------------------------------------------------------------------------------+
190+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_8 OnXpuProfiling RUNNING 0s |
191++---------------------------------------------------------------------------------------------------------+
192+2026-09-28 01:33:47,936 [INFO] [1673932 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
193+2026-09-28 01:33:47,936 [INFO] [1673932 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_8.json
194+2026-09-28 01:33:47,936 [INFO] [1673932 auto_testcase_name_8 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
195+2026-09-28 01:33:53,583 [INFO] [1673918 MP MT] [instance_base.py:301]:
196++---------------------------------------------------------------------------------------------------------+
197+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 150s Rev: 9bfa978 |
198++---------------------------------------------------------------------------------------------------------+
199+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_8 OnGeirExecute RUNNING 5s |
200++---------------------------------------------------------------------------------------------------------+
201+2026-09-28 01:34:03,589 [INFO] [1673918 MP MT] [instance_base.py:301]:
202++----------------------------------------------------------------------------------------------------------+
203+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 160s Rev: 9bfa978 |
204++----------------------------------------------------------------------------------------------------------+
205+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_8 OnGeirExecute RUNNING 15s |
206++----------------------------------------------------------------------------------------------------------+
207+2026-09-28 01:34:03,713 [INFO] [1673932 auto_testcase_name_8 OnReturning] [profiling.py:63]:
208+########################
209+Performance result: Comparison result:
210+CST_PERF: None CST_GOLD: PASS,PASS
211+DYN_PERF: None DYN_GOLD:
212+PRECISION_STATUS: PASS PASSED: PASS
213+########################
214+
215+2026-09-28 01:34:08,235 [INFO] [1673932 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
216+2026-09-28 01:34:08,235 [INFO] [1673932 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_9.json
217+2026-09-28 01:34:08,236 [INFO] [1673932 auto_testcase_name_9 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
218+2026-09-28 01:34:13,598 [INFO] [1673918 MP MT] [instance_base.py:301]:
219++---------------------------------------------------------------------------------------------------------+
220+| Version: 3.0.0 Summary (Device Total: 1) Progress: 80% 8 / 10 ET: 170s Rev: 9bfa978 |
221++---------------------------------------------------------------------------------------------------------+
222+| NPU:0 Ascend950PR_9579 | 1673932 _testcase_name_9 OnGeirExecute RUNNING 5s |
223++---------------------------------------------------------------------------------------------------------+
224+2026-09-28 01:34:22,907 [INFO] [1673932 auto_testcase_name_9 OnReturning] [profiling.py:63]:
225+########################
226+Performance result: Comparison result:
227+CST_PERF: None CST_GOLD: PASS,PASS
228+DYN_PERF: None DYN_GOLD:
229+PRECISION_STATUS: PASS PASSED: PASS
230+########################
231+
232+2026-09-28 01:34:23,606 [INFO] [1673918 MP MT] [instance_base.py:301]:
233++---------------------------------------------------------------------------------------------------------+
234+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 180s Rev: 9bfa978 |
235++---------------------------------------------------------------------------------------------------------+
236+| NPU:0 Ascend950PR_9579 | 1673932 testcase_name_10 OnXpuProfiling RUNNING 0s |
237++---------------------------------------------------------------------------------------------------------+
238+2026-09-28 01:34:27,435 [INFO] [1673932 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
239+2026-09-28 01:34:27,435 [INFO] [1673932 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_10.json
240+2026-09-28 01:34:27,436 [INFO] [1673932 auto_testcase_name_10 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
241+2026-09-28 01:34:33,614 [INFO] [1673918 MP MT] [instance_base.py:301]:
242++---------------------------------------------------------------------------------------------------------+
243+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 190s Rev: 9bfa978 |
244++---------------------------------------------------------------------------------------------------------+
245+| NPU:0 Ascend950PR_9579 | 1673932 testcase_name_10 OnGeirExecute RUNNING 6s |
246++---------------------------------------------------------------------------------------------------------+
247+2026-09-28 01:34:41,601 [INFO] [1673932 auto_testcase_name_10 OnReturning] [profiling.py:63]:
248+########################
249+Performance result: Comparison result:
250+CST_PERF: None CST_GOLD: PASS,PASS
251+DYN_PERF: None DYN_GOLD:
252+PRECISION_STATUS: PASS PASSED: PASS
253+########################
254+
255+2026-09-28 01:34:41,618 [INFO] [1673918 MP MT] [instance_base.py:147]: ttk Profiling complete
256+2026-09-28 01:34:41,689 [INFO] [1673918 MP MT] [heartbeat_manager.py:158]: tenant 1d8c46ecacc1 deleted on 127.0.0.1:19098
257+2026-09-28 01:34:41,697 [INFO] [1673918 MP MT] [instance_base.py:325]:
258+==================== TTK Test Summary ===================
259+ Total : 10
260+ PASS : 10
261+ FAIL : 0
262+ SKIP/NA : 0
263+ PassRate : 100.00% (10/10)
264+=========================================================
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_01.csv+11-0
@@ -0,0 +1,11 @@
1+testcase_name,network_name,op_name,input_shapes,input_dtypes,input_formats,output_shapes,output_dtypes,output_formats,input_ori_shapes,input_ori_formats,output_ori_shapes,output_ori_formats,attributes,input_data_ranges,precision_tolerances,absolute_precision,output_inplace_indexes,output_shape_unknown_indexes,is_enabled,remark,soc_series,priority,dump_file_prefix,manual_input_binaries,manual_golden_binaries
2+in_training_update_grad_gamma_beta_gen0204_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[1, 31, 8, 65], [1, 31, 8, 65]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 31, 8, 65], [1, 31, 8, 65]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 31, 8, 65], [1, 31, 8, 65]]","['NCHW', 'NCHW']","[[1, 31, 8, 65], [1, 31, 8, 65]]","['NCHW', 'NCHW']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
3+in_training_update_grad_gamma_beta_gen0224_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[2, 1, 16, 648], [2, 1, 16, 648]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 1, 16, 648], [1, 1, 16, 648]]","['float32', 'float32']","['NCHW', 'NCHW']","[[2, 1, 16, 648], [2, 1, 16, 648]]","['NCHW', 'NCHW']","[[1, 1, 16, 648], [1, 1, 16, 648]]","['NCHW', 'NCHW']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
4+in_training_update_grad_gamma_beta_gen0245_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[63, 16, 5, 2], [63, 16, 5, 2]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 16, 5, 2], [1, 16, 5, 2]]","['float32', 'float32']","['NHWC', 'NHWC']","[[63, 16, 5, 2], [63, 16, 5, 2]]","['NHWC', 'NHWC']","[[1, 16, 5, 2], [1, 16, 5, 2]]","['NHWC', 'NHWC']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
5+in_training_update_grad_gamma_beta_gen0265_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[3, 4, 2, 4], [3, 4, 2, 4]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 4, 2, 4], [1, 4, 2, 4]]","['float32', 'float32']","['NHWC', 'NHWC']","[[3, 4, 2, 4], [3, 4, 2, 4]]","['NHWC', 'NHWC']","[[1, 4, 2, 4], [1, 4, 2, 4]]","['NHWC', 'NHWC']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
6+in_training_update_grad_gamma_beta_gen0285_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[64, 1, 31, 5], [64, 1, 31, 5]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 1, 31, 5], [1, 1, 31, 5]]","['float32', 'float32']","['NHWC', 'NHWC']","[[64, 1, 31, 5], [64, 1, 31, 5]]","['NHWC', 'NHWC']","[[1, 1, 31, 5], [1, 1, 31, 5]]","['NHWC', 'NHWC']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
7+in_training_update_grad_gamma_beta_gen0306_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[4, 1, 3, 1], [4, 1, 3, 1]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 1, 3, 1], [1, 1, 3, 1]]","['float32', 'float32']","['NCHW', 'NCHW']","[[4, 1, 3, 1], [4, 1, 3, 1]]","['NCHW', 'NCHW']","[[1, 1, 3, 1], [1, 1, 3, 1]]","['NCHW', 'NCHW']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
8+in_training_update_grad_gamma_beta_gen0326_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[3, 31, 5, 65], [3, 31, 5, 65]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 31, 5, 65], [1, 31, 5, 65]]","['float32', 'float32']","['NCHW', 'NCHW']","[[3, 31, 5, 65], [3, 31, 5, 65]]","['NCHW', 'NCHW']","[[1, 31, 5, 65], [1, 31, 5, 65]]","['NCHW', 'NCHW']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
9+in_training_update_grad_gamma_beta_gen0346_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[31, 2, 32, 31], [31, 2, 32, 31]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 2, 32, 31], [1, 2, 32, 31]]","['float32', 'float32']","['NCHW', 'NCHW']","[[31, 2, 32, 31], [31, 2, 32, 31]]","['NCHW', 'NCHW']","[[1, 2, 32, 31], [1, 2, 32, 31]]","['NCHW', 'NCHW']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
10+in_training_update_grad_gamma_beta_gen0367_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[8, 16, 7, 3], [8, 16, 7, 3]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 16, 7, 3], [1, 16, 7, 3]]","['float32', 'float32']","['NHWC', 'NHWC']","[[8, 16, 7, 3], [8, 16, 7, 3]]","['NHWC', 'NHWC']","[[1, 16, 7, 3], [1, 16, 7, 3]]","['NHWC', 'NHWC']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
11+in_training_update_grad_gamma_beta_gen0387_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[257, 16, 16, 15], [257, 16, 16, 15]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 16, 16, 15], [1, 16, 16, 15]]","['float32', 'float32']","['NHWC', 'NHWC']","[[257, 16, 16, 15], [257, 16, 16, 15]]","['NHWC', 'NHWC']","[[1, 16, 16, 15], [1, 16, 16, 15]]","['NHWC', 'NHWC']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_01_run.log+264-0
@@ -0,0 +1,264 @@
1+2026-09-28 02:09:00,391 [DEBUG] [1754534 MainProcess MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
2+2026-09-28 02:09:00,393 [INFO] [1754534 MP MT] [bridge.py:441]: Command: ttk geir -i /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_01.csv
3+2026-09-28 02:09:00,394 [INFO] [1754534 MP MT] [instance_base.py:118]: TTK Profiling Start!
4+2026-09-28 02:09:00,394 [INFO] [1754534 MP MT] [instance_base.py:119]: Mode: ASCEND_ONBOARD
5+2026-09-28 02:09:00,398 [INFO] [1754534 MP MT] [instance_base.py:122]: Device Platform: Ascend950PR_9579
6+2026-09-28 02:09:00,398 [INFO] [1754534 MP MT] [instance_base.py:129]: Device Count: 1
7+2026-09-28 02:09:00,398 [INFO] [1754534 MP MT] [instance_base.py:372]: Preparing testcases ...
8+2026-09-28 02:09:00,398 [INFO] [1754534 MP MT] [instance_base.py:374]: Reading testcase file /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_01.csv ...
9+2026-09-28 02:09:00,398 [INFO] [1754534 MP MT] [instance_base.py:447]: Input testcase file is not a valid zip file, switch to normal csv mode...
10+2026-09-28 02:09:00,399 [INFO] [1754534 MP MT] [instance_base.py:723]: Reading testcases from /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_01.csv ...
11+2026-09-28 02:09:00,399 [WARNING] [1754534 TestcaseManager HeaderCheckTestcaseName] [testcase_manager.py:258]: Testcase name not found! It is important to add a testcase_name in order to identify your testcases
12+2026-09-28 02:09:00,399 [WARNING] [1754534 TestcaseManager HeaderCheckUnidentifiedHeaders] [testcase_manager.py:289]: Detected unidentified headers and will be ignored: ['\ufefftestcase_name']
13+2026-09-28 02:09:00,402 [INFO] [1754534 TestcaseManager auto_testcase_name_1] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
14+2026-09-28 02:09:00,455 [INFO] [1754534 MP MT] [instance_base.py:378]: Checking testcase name...
15+2026-09-28 02:09:00,455 [INFO] [1754534 MP MT] [instance_base.py:381]: Case number: 10
16+2026-09-28 02:09:00,467 [INFO] [1754534 MP MT] [heartbeat_manager.py:86]: Heartbeat subprocess started: pid=1754547, health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_61f972ef0eb1.json
17+2026-09-28 02:09:00,470 [INFO] [1754547 HB MT] [heartbeat.py:50]: HB started: tenant=61f972ef0eb1 health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_61f972ef0eb1.json n_endpoints=1 tls=False
18+2026-09-28 02:09:00,507 [INFO] [1754534 MP MT] [instance_base.py:463]: Initialize output csv file...
19+2026-09-28 02:09:00,551 [INFO] [1754534 MP MT] [instance_base.py:167]: Process per device: 1
20+2026-09-28 02:09:00,551 [INFO] [1754534 MP MT] [instance_base.py:169]: Preparing Task Executors...
21+2026-09-28 02:09:00,649 [DEBUG] [1754559 D0P0 MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
22+2026-09-28 02:09:00,656 [INFO] [1754534 MP MT] [instance_base.py:301]:
23++--------------------------------------------------------------------------------------+
24+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 0s Rev: 9bfa978 |
25++--------------------------------------------------------------------------------------+
26+| NPU:0 Ascend950PR_9579 | 1754559 D0P0 UNKNOWN IDLE 0s |
27++--------------------------------------------------------------------------------------+
28+2026-09-28 02:09:00,693 [INFO] [1754559 auto_testcase_name_1 OnGenerateInput] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
29+2026-09-28 02:09:03,456 [WARNING] [1754559 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:550]: 424 retry (1/5): {"error": "missing spec dependency: golden", "missing": "golden"}
30+2026-09-28 02:09:03,540 [INFO] [1754559 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:357]: 424: synced 'golden.py' for missing module 'golden'
31+2026-09-28 02:09:07,919 [INFO] [1754559 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
32+2026-09-28 02:09:07,919 [INFO] [1754559 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_1.json
33+2026-09-28 02:09:07,919 [INFO] [1754559 auto_testcase_name_1 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
34+2026-09-28 02:09:10,655 [INFO] [1754534 MP MT] [instance_base.py:301]:
35++--------------------------------------------------------------------------------------+
36+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 10s Rev: 9bfa978 |
37++--------------------------------------------------------------------------------------+
38+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_1 OnGeirExecute RUNNING 2s |
39++--------------------------------------------------------------------------------------+
40+2026-09-28 02:09:20,662 [INFO] [1754534 MP MT] [instance_base.py:301]:
41++---------------------------------------------------------------------------------------+
42+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 20s Rev: 9bfa978 |
43++---------------------------------------------------------------------------------------+
44+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_1 OnGeirExecute RUNNING 12s |
45++---------------------------------------------------------------------------------------+
46+2026-09-28 02:09:23,078 [INFO] [1754559 auto_testcase_name_1 OnReturning] [profiling.py:63]:
47+########################
48+Performance result: Comparison result:
49+CST_PERF: None CST_GOLD: PASS,PASS
50+DYN_PERF: None DYN_GOLD:
51+PRECISION_STATUS: PASS PASSED: PASS
52+########################
53+
54+2026-09-28 02:09:27,621 [INFO] [1754559 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
55+2026-09-28 02:09:27,621 [INFO] [1754559 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_2.json
56+2026-09-28 02:09:27,621 [INFO] [1754559 auto_testcase_name_2 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
57+2026-09-28 02:09:30,664 [INFO] [1754534 MP MT] [instance_base.py:301]:
58++--------------------------------------------------------------------------------------------------------+
59+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 30s Rev: 9bfa978 |
60++--------------------------------------------------------------------------------------------------------+
61+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_2 OnGeirExecute RUNNING 3s |
62++--------------------------------------------------------------------------------------------------------+
63+2026-09-28 02:09:40,667 [INFO] [1754534 MP MT] [instance_base.py:301]:
64++---------------------------------------------------------------------------------------+
65+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 40s Rev: 9bfa978 |
66++---------------------------------------------------------------------------------------+
67+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_2 OnGeirExecute RUNNING 13s |
68++---------------------------------------------------------------------------------------+
69+2026-09-28 02:09:42,615 [INFO] [1754559 auto_testcase_name_2 OnReturning] [profiling.py:63]:
70+########################
71+Performance result: Comparison result:
72+CST_PERF: None CST_GOLD: PASS,PASS
73+DYN_PERF: None DYN_GOLD:
74+PRECISION_STATUS: PASS PASSED: PASS
75+########################
76+
77+2026-09-28 02:09:47,083 [INFO] [1754559 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
78+2026-09-28 02:09:47,084 [INFO] [1754559 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_3.json
79+2026-09-28 02:09:47,084 [INFO] [1754559 auto_testcase_name_3 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
80+2026-09-28 02:09:50,674 [INFO] [1754534 MP MT] [instance_base.py:301]:
81++--------------------------------------------------------------------------------------------------------+
82+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 50s Rev: 9bfa978 |
83++--------------------------------------------------------------------------------------------------------+
84+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_3 OnGeirExecute RUNNING 3s |
85++--------------------------------------------------------------------------------------------------------+
86+2026-09-28 02:10:00,679 [INFO] [1754534 MP MT] [instance_base.py:301]:
87++---------------------------------------------------------------------------------------+
88+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 60s Rev: 9bfa978 |
89++---------------------------------------------------------------------------------------+
90+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_3 OnGeirExecute RUNNING 13s |
91++---------------------------------------------------------------------------------------+
92+2026-09-28 02:10:01,961 [INFO] [1754559 auto_testcase_name_3 OnReturning] [profiling.py:63]:
93+########################
94+Performance result: Comparison result:
95+CST_PERF: None CST_GOLD: PASS,PASS
96+DYN_PERF: None DYN_GOLD:
97+PRECISION_STATUS: PASS PASSED: PASS
98+########################
99+
100+2026-09-28 02:10:06,486 [INFO] [1754559 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
101+2026-09-28 02:10:06,487 [INFO] [1754559 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_4.json
102+2026-09-28 02:10:06,487 [INFO] [1754559 auto_testcase_name_4 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
103+2026-09-28 02:10:10,686 [INFO] [1754534 MP MT] [instance_base.py:301]:
104++--------------------------------------------------------------------------------------------------------+
105+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 70s Rev: 9bfa978 |
106++--------------------------------------------------------------------------------------------------------+
107+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_4 OnGeirExecute RUNNING 4s |
108++--------------------------------------------------------------------------------------------------------+
109+2026-09-28 02:10:20,691 [INFO] [1754534 MP MT] [instance_base.py:301]:
110++---------------------------------------------------------------------------------------+
111+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 80s Rev: 9bfa978 |
112++---------------------------------------------------------------------------------------+
113+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_4 OnGeirExecute RUNNING 14s |
114++---------------------------------------------------------------------------------------+
115+2026-09-28 02:10:21,246 [INFO] [1754559 auto_testcase_name_4 OnReturning] [profiling.py:63]:
116+########################
117+Performance result: Comparison result:
118+CST_PERF: None CST_GOLD: PASS,PASS
119+DYN_PERF: None DYN_GOLD:
120+PRECISION_STATUS: PASS PASSED: PASS
121+########################
122+
123+2026-09-28 02:10:25,849 [INFO] [1754559 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
124+2026-09-28 02:10:25,849 [INFO] [1754559 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_5.json
125+2026-09-28 02:10:25,849 [INFO] [1754559 auto_testcase_name_5 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
126+2026-09-28 02:10:30,697 [INFO] [1754534 MP MT] [instance_base.py:301]:
127++--------------------------------------------------------------------------------------------------------+
128+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 90s Rev: 9bfa978 |
129++--------------------------------------------------------------------------------------------------------+
130+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_5 OnGeirExecute RUNNING 4s |
131++--------------------------------------------------------------------------------------------------------+
132+2026-09-28 02:10:40,553 [INFO] [1754559 auto_testcase_name_5 OnReturning] [profiling.py:63]:
133+########################
134+Performance result: Comparison result:
135+CST_PERF: None CST_GOLD: PASS,PASS
136+DYN_PERF: None DYN_GOLD:
137+PRECISION_STATUS: PASS PASSED: PASS
138+########################
139+
140+2026-09-28 02:10:40,707 [INFO] [1754534 MP MT] [instance_base.py:301]:
141++---------------------------------------------------------------------------------------------------------+
142+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 100s Rev: 9bfa978 |
143++---------------------------------------------------------------------------------------------------------+
144+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_6 OnXpuProfiling RUNNING 0s |
145++---------------------------------------------------------------------------------------------------------+
146+2026-09-28 02:10:45,080 [INFO] [1754559 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
147+2026-09-28 02:10:45,081 [INFO] [1754559 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_6.json
148+2026-09-28 02:10:45,081 [INFO] [1754559 auto_testcase_name_6 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
149+2026-09-28 02:10:50,713 [INFO] [1754534 MP MT] [instance_base.py:301]:
150++---------------------------------------------------------------------------------------------------------+
151+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 110s Rev: 9bfa978 |
152++---------------------------------------------------------------------------------------------------------+
153+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_6 OnGeirExecute RUNNING 5s |
154++---------------------------------------------------------------------------------------------------------+
155+2026-09-28 02:11:00,722 [INFO] [1754534 MP MT] [instance_base.py:301]:
156++----------------------------------------------------------------------------------------------------------+
157+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 120s Rev: 9bfa978 |
158++----------------------------------------------------------------------------------------------------------+
159+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_6 OnGeirExecute RUNNING 15s |
160++----------------------------------------------------------------------------------------------------------+
161+2026-09-28 02:11:00,982 [INFO] [1754559 auto_testcase_name_6 OnReturning] [profiling.py:63]:
162+########################
163+Performance result: Comparison result:
164+CST_PERF: None CST_GOLD: PASS,PASS
165+DYN_PERF: None DYN_GOLD:
166+PRECISION_STATUS: PASS PASSED: PASS
167+########################
168+
169+2026-09-28 02:11:05,516 [INFO] [1754559 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
170+2026-09-28 02:11:05,516 [INFO] [1754559 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_7.json
171+2026-09-28 02:11:05,517 [INFO] [1754559 auto_testcase_name_7 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
172+2026-09-28 02:11:10,732 [INFO] [1754534 MP MT] [instance_base.py:301]:
173++---------------------------------------------------------------------------------------------------------+
174+| Version: 3.0.0 Summary (Device Total: 1) Progress: 60% 6 / 10 ET: 130s Rev: 9bfa978 |
175++---------------------------------------------------------------------------------------------------------+
176+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_7 OnGeirExecute RUNNING 5s |
177++---------------------------------------------------------------------------------------------------------+
178+2026-09-28 02:11:20,328 [INFO] [1754559 auto_testcase_name_7 OnReturning] [profiling.py:63]:
179+########################
180+Performance result: Comparison result:
181+CST_PERF: None CST_GOLD: PASS,PASS
182+DYN_PERF: None DYN_GOLD:
183+PRECISION_STATUS: PASS PASSED: PASS
184+########################
185+
186+2026-09-28 02:11:20,740 [INFO] [1754534 MP MT] [instance_base.py:301]:
187++---------------------------------------------------------------------------------------------------------+
188+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 140s Rev: 9bfa978 |
189++---------------------------------------------------------------------------------------------------------+
190+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_8 OnXpuProfiling RUNNING 0s |
191++---------------------------------------------------------------------------------------------------------+
192+2026-09-28 02:11:24,849 [INFO] [1754559 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
193+2026-09-28 02:11:24,850 [INFO] [1754559 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_8.json
194+2026-09-28 02:11:24,850 [INFO] [1754559 auto_testcase_name_8 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
195+2026-09-28 02:11:30,745 [INFO] [1754534 MP MT] [instance_base.py:301]:
196++---------------------------------------------------------------------------------------------------------+
197+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 150s Rev: 9bfa978 |
198++---------------------------------------------------------------------------------------------------------+
199+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_8 OnGeirExecute RUNNING 5s |
200++---------------------------------------------------------------------------------------------------------+
201+2026-09-28 02:11:39,574 [INFO] [1754559 auto_testcase_name_8 OnReturning] [profiling.py:63]:
202+########################
203+Performance result: Comparison result:
204+CST_PERF: None CST_GOLD: PASS,PASS
205+DYN_PERF: None DYN_GOLD:
206+PRECISION_STATUS: PASS PASSED: PASS
207+########################
208+
209+2026-09-28 02:11:40,751 [INFO] [1754534 MP MT] [instance_base.py:301]:
210++---------------------------------------------------------------------------------------------------------+
211+| Version: 3.0.0 Summary (Device Total: 1) Progress: 80% 8 / 10 ET: 160s Rev: 9bfa978 |
212++---------------------------------------------------------------------------------------------------------+
213+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_9 OnXpuProfiling RUNNING 1s |
214++---------------------------------------------------------------------------------------------------------+
215+2026-09-28 02:11:44,076 [INFO] [1754559 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
216+2026-09-28 02:11:44,077 [INFO] [1754559 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_9.json
217+2026-09-28 02:11:44,077 [INFO] [1754559 auto_testcase_name_9 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
218+2026-09-28 02:11:50,761 [INFO] [1754534 MP MT] [instance_base.py:301]:
219++---------------------------------------------------------------------------------------------------------+
220+| Version: 3.0.0 Summary (Device Total: 1) Progress: 80% 8 / 10 ET: 170s Rev: 9bfa978 |
221++---------------------------------------------------------------------------------------------------------+
222+| NPU:0 Ascend950PR_9579 | 1754559 _testcase_name_9 OnGeirExecute RUNNING 6s |
223++---------------------------------------------------------------------------------------------------------+
224+2026-09-28 02:11:58,656 [INFO] [1754559 auto_testcase_name_9 OnReturning] [profiling.py:63]:
225+########################
226+Performance result: Comparison result:
227+CST_PERF: None CST_GOLD: PASS,PASS
228+DYN_PERF: None DYN_GOLD:
229+PRECISION_STATUS: PASS PASSED: PASS
230+########################
231+
232+2026-09-28 02:12:00,761 [INFO] [1754534 MP MT] [instance_base.py:301]:
233++---------------------------------------------------------------------------------------------------------+
234+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 180s Rev: 9bfa978 |
235++---------------------------------------------------------------------------------------------------------+
236+| NPU:0 Ascend950PR_9579 | 1754559 testcase_name_10 OnXpuProfiling RUNNING 2s |
237++---------------------------------------------------------------------------------------------------------+
238+2026-09-28 02:12:03,396 [INFO] [1754559 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
239+2026-09-28 02:12:03,397 [INFO] [1754559 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_10.json
240+2026-09-28 02:12:03,397 [INFO] [1754559 auto_testcase_name_10 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
241+2026-09-28 02:12:10,768 [INFO] [1754534 MP MT] [instance_base.py:301]:
242++---------------------------------------------------------------------------------------------------------+
243+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 190s Rev: 9bfa978 |
244++---------------------------------------------------------------------------------------------------------+
245+| NPU:0 Ascend950PR_9579 | 1754559 testcase_name_10 OnGeirExecute RUNNING 7s |
246++---------------------------------------------------------------------------------------------------------+
247+2026-09-28 02:12:18,162 [INFO] [1754559 auto_testcase_name_10 OnReturning] [profiling.py:63]:
248+########################
249+Performance result: Comparison result:
250+CST_PERF: None CST_GOLD: PASS,PASS
251+DYN_PERF: None DYN_GOLD:
252+PRECISION_STATUS: PASS PASSED: PASS
253+########################
254+
255+2026-09-28 02:12:18,184 [INFO] [1754534 MP MT] [instance_base.py:147]: ttk Profiling complete
256+2026-09-28 02:12:18,209 [INFO] [1754534 MP MT] [heartbeat_manager.py:158]: tenant 61f972ef0eb1 deleted on 127.0.0.1:19098
257+2026-09-28 02:12:18,212 [INFO] [1754534 MP MT] [instance_base.py:325]:
258+==================== TTK Test Summary ===================
259+ Total : 10
260+ PASS : 10
261+ FAIL : 0
262+ SKIP/NA : 0
263+ PassRate : 100.00% (10/10)
264+=========================================================
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_02.csv+11-0
@@ -0,0 +1,11 @@
1+testcase_name,network_name,op_name,input_shapes,input_dtypes,input_formats,output_shapes,output_dtypes,output_formats,input_ori_shapes,input_ori_formats,output_ori_shapes,output_ori_formats,attributes,input_data_ranges,precision_tolerances,absolute_precision,output_inplace_indexes,output_shape_unknown_indexes,is_enabled,remark,soc_series,priority,dump_file_prefix,manual_input_binaries,manual_golden_binaries
2+in_training_update_grad_gamma_beta_gen0407_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[127, 2, 7, 257], [127, 2, 7, 257]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 2, 7, 257], [1, 2, 7, 257]]","['float32', 'float32']","['NHWC', 'NHWC']","[[127, 2, 7, 257], [127, 2, 7, 257]]","['NHWC', 'NHWC']","[[1, 2, 7, 257], [1, 2, 7, 257]]","['NHWC', 'NHWC']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
3+in_training_update_grad_gamma_beta_gen0427_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[5, 31, 7, 32], [5, 31, 7, 32]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 31, 7, 32], [1, 31, 7, 32]]","['float32', 'float32']","['NHWC', 'NHWC']","[[5, 31, 7, 32], [5, 31, 7, 32]]","['NHWC', 'NHWC']","[[1, 31, 7, 32], [1, 31, 7, 32]]","['NHWC', 'NHWC']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
4+in_training_update_grad_gamma_beta_gen0449_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[3, 4, 32, 647], [3, 4, 32, 647]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 4, 32, 647], [1, 4, 32, 647]]","['float32', 'float32']","['NHWC', 'NHWC']","[[3, 4, 32, 647], [3, 4, 32, 647]]","['NHWC', 'NHWC']","[[1, 4, 32, 647], [1, 4, 32, 647]]","['NHWC', 'NHWC']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
5+in_training_update_grad_gamma_beta_gen0469_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[63, 4, 15, 16], [63, 4, 15, 16]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 4, 15, 16], [1, 4, 15, 16]]","['float32', 'float32']","['NHWC', 'NHWC']","[[63, 4, 15, 16], [63, 4, 15, 16]]","['NHWC', 'NHWC']","[[1, 4, 15, 16], [1, 4, 15, 16]]","['NHWC', 'NHWC']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
6+in_training_update_grad_gamma_beta_gen0489_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[5, 31, 5, 3], [5, 31, 5, 3]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 31, 5, 3], [1, 31, 5, 3]]","['float32', 'float32']","['NHWC', 'NHWC']","[[5, 31, 5, 3], [5, 31, 5, 3]]","['NHWC', 'NHWC']","[[1, 31, 5, 3], [1, 31, 5, 3]]","['NHWC', 'NHWC']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
7+in_training_update_grad_gamma_beta_gen0510_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[513, 7, 2, 127], [513, 7, 2, 127]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 7, 2, 127], [1, 7, 2, 127]]","['float32', 'float32']","['NCHW', 'NCHW']","[[513, 7, 2, 127], [513, 7, 2, 127]]","['NCHW', 'NCHW']","[[1, 7, 2, 127], [1, 7, 2, 127]]","['NCHW', 'NCHW']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
8+in_training_update_grad_gamma_beta_gen0530_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[4, 3, 31, 257], [4, 3, 31, 257]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 3, 31, 257], [1, 3, 31, 257]]","['float32', 'float32']","['NCHW', 'NCHW']","[[4, 3, 31, 257], [4, 3, 31, 257]]","['NCHW', 'NCHW']","[[1, 3, 31, 257], [1, 3, 31, 257]]","['NCHW', 'NCHW']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
9+in_training_update_grad_gamma_beta_gen0550_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[127, 2, 1, 127], [127, 2, 1, 127]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 2, 1, 127], [1, 2, 1, 127]]","['float32', 'float32']","['NCHW', 'NCHW']","[[127, 2, 1, 127], [127, 2, 1, 127]]","['NCHW', 'NCHW']","[[1, 2, 1, 127], [1, 2, 1, 127]]","['NCHW', 'NCHW']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
10+in_training_update_grad_gamma_beta_gen0571_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[3, 4, 3, 15], [3, 4, 3, 15]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 4, 3, 15], [1, 4, 3, 15]]","['float32', 'float32']","['NHWC', 'NHWC']","[[3, 4, 3, 15], [3, 4, 3, 15]]","['NHWC', 'NHWC']","[[1, 4, 3, 15], [1, 4, 3, 15]]","['NHWC', 'NHWC']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
11+in_training_update_grad_gamma_beta_gen0591_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[2, 7, 4, 65], [2, 7, 4, 65]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 7, 4, 65], [1, 7, 4, 65]]","['float32', 'float32']","['NHWC', 'NHWC']","[[2, 7, 4, 65], [2, 7, 4, 65]]","['NHWC', 'NHWC']","[[1, 7, 4, 65], [1, 7, 4, 65]]","['NHWC', 'NHWC']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_02_run.log+264-0
@@ -0,0 +1,264 @@
1+2026-09-28 02:14:39,574 [DEBUG] [1769061 MainProcess MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
2+2026-09-28 02:14:39,576 [INFO] [1769061 MP MT] [bridge.py:441]: Command: ttk geir -i /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_02.csv
3+2026-09-28 02:14:39,577 [INFO] [1769061 MP MT] [instance_base.py:118]: TTK Profiling Start!
4+2026-09-28 02:14:39,577 [INFO] [1769061 MP MT] [instance_base.py:119]: Mode: ASCEND_ONBOARD
5+2026-09-28 02:14:39,581 [INFO] [1769061 MP MT] [instance_base.py:122]: Device Platform: Ascend950PR_9579
6+2026-09-28 02:14:39,581 [INFO] [1769061 MP MT] [instance_base.py:129]: Device Count: 1
7+2026-09-28 02:14:39,581 [INFO] [1769061 MP MT] [instance_base.py:372]: Preparing testcases ...
8+2026-09-28 02:14:39,581 [INFO] [1769061 MP MT] [instance_base.py:374]: Reading testcase file /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_02.csv ...
9+2026-09-28 02:14:39,581 [INFO] [1769061 MP MT] [instance_base.py:447]: Input testcase file is not a valid zip file, switch to normal csv mode...
10+2026-09-28 02:14:39,581 [INFO] [1769061 MP MT] [instance_base.py:723]: Reading testcases from /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_02.csv ...
11+2026-09-28 02:14:39,581 [WARNING] [1769061 TestcaseManager HeaderCheckTestcaseName] [testcase_manager.py:258]: Testcase name not found! It is important to add a testcase_name in order to identify your testcases
12+2026-09-28 02:14:39,582 [WARNING] [1769061 TestcaseManager HeaderCheckUnidentifiedHeaders] [testcase_manager.py:289]: Detected unidentified headers and will be ignored: ['\ufefftestcase_name']
13+2026-09-28 02:14:39,585 [INFO] [1769061 TestcaseManager auto_testcase_name_1] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
14+2026-09-28 02:14:39,637 [INFO] [1769061 MP MT] [instance_base.py:378]: Checking testcase name...
15+2026-09-28 02:14:39,637 [INFO] [1769061 MP MT] [instance_base.py:381]: Case number: 10
16+2026-09-28 02:14:39,649 [INFO] [1769061 MP MT] [heartbeat_manager.py:86]: Heartbeat subprocess started: pid=1769070, health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_4c9186aa8151.json
17+2026-09-28 02:14:39,652 [INFO] [1769070 HB MT] [heartbeat.py:50]: HB started: tenant=4c9186aa8151 health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_4c9186aa8151.json n_endpoints=1 tls=False
18+2026-09-28 02:14:39,687 [INFO] [1769061 MP MT] [instance_base.py:463]: Initialize output csv file...
19+2026-09-28 02:14:39,731 [INFO] [1769061 MP MT] [instance_base.py:167]: Process per device: 1
20+2026-09-28 02:14:39,731 [INFO] [1769061 MP MT] [instance_base.py:169]: Preparing Task Executors...
21+2026-09-28 02:14:39,828 [DEBUG] [1769082 D0P0 MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
22+2026-09-28 02:14:39,836 [INFO] [1769061 MP MT] [instance_base.py:301]:
23++--------------------------------------------------------------------------------------+
24+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 0s Rev: 9bfa978 |
25++--------------------------------------------------------------------------------------+
26+| NPU:0 Ascend950PR_9579 | 1769082 D0P0 UNKNOWN IDLE 0s |
27++--------------------------------------------------------------------------------------+
28+2026-09-28 02:14:39,878 [INFO] [1769082 auto_testcase_name_1 OnGenerateInput] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
29+2026-09-28 02:14:42,706 [WARNING] [1769082 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:550]: 424 retry (1/5): {"error": "missing spec dependency: golden", "missing": "golden"}
30+2026-09-28 02:14:42,780 [INFO] [1769082 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:357]: 424: synced 'golden.py' for missing module 'golden'
31+2026-09-28 02:14:49,185 [INFO] [1769082 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
32+2026-09-28 02:14:49,186 [INFO] [1769082 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_1.json
33+2026-09-28 02:14:49,186 [INFO] [1769082 auto_testcase_name_1 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
34+2026-09-28 02:14:49,835 [INFO] [1769061 MP MT] [instance_base.py:301]:
35++--------------------------------------------------------------------------------------+
36+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 10s Rev: 9bfa978 |
37++--------------------------------------------------------------------------------------+
38+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_1 OnGeirExecute RUNNING 0s |
39++--------------------------------------------------------------------------------------+
40+2026-09-28 02:14:59,842 [INFO] [1769061 MP MT] [instance_base.py:301]:
41++---------------------------------------------------------------------------------------+
42+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 20s Rev: 9bfa978 |
43++---------------------------------------------------------------------------------------+
44+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_1 OnGeirExecute RUNNING 10s |
45++---------------------------------------------------------------------------------------+
46+2026-09-28 02:15:04,640 [INFO] [1769082 auto_testcase_name_1 OnReturning] [profiling.py:63]:
47+########################
48+Performance result: Comparison result:
49+CST_PERF: None CST_GOLD: PASS,PASS
50+DYN_PERF: None DYN_GOLD:
51+PRECISION_STATUS: PASS PASSED: PASS
52+########################
53+
54+2026-09-28 02:15:09,238 [INFO] [1769082 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
55+2026-09-28 02:15:09,239 [INFO] [1769082 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_2.json
56+2026-09-28 02:15:09,239 [INFO] [1769082 auto_testcase_name_2 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
57+2026-09-28 02:15:09,842 [INFO] [1769061 MP MT] [instance_base.py:301]:
58++--------------------------------------------------------------------------------------------------------+
59+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 30s Rev: 9bfa978 |
60++--------------------------------------------------------------------------------------------------------+
61+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_2 OnGeirExecute RUNNING 0s |
62++--------------------------------------------------------------------------------------------------------+
63+2026-09-28 02:15:19,851 [INFO] [1769061 MP MT] [instance_base.py:301]:
64++---------------------------------------------------------------------------------------+
65+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 40s Rev: 9bfa978 |
66++---------------------------------------------------------------------------------------+
67+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_2 OnGeirExecute RUNNING 10s |
68++---------------------------------------------------------------------------------------+
69+2026-09-28 02:15:24,559 [INFO] [1769082 auto_testcase_name_2 OnReturning] [profiling.py:63]:
70+########################
71+Performance result: Comparison result:
72+CST_PERF: None CST_GOLD: PASS,PASS
73+DYN_PERF: None DYN_GOLD:
74+PRECISION_STATUS: PASS PASSED: PASS
75+########################
76+
77+2026-09-28 02:15:29,351 [INFO] [1769082 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
78+2026-09-28 02:15:29,351 [INFO] [1769082 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_3.json
79+2026-09-28 02:15:29,351 [INFO] [1769082 auto_testcase_name_3 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
80+2026-09-28 02:15:29,859 [INFO] [1769061 MP MT] [instance_base.py:301]:
81++--------------------------------------------------------------------------------------------------------+
82+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 50s Rev: 9bfa978 |
83++--------------------------------------------------------------------------------------------------------+
84+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_3 OnGeirExecute RUNNING 0s |
85++--------------------------------------------------------------------------------------------------------+
86+2026-09-28 02:15:39,867 [INFO] [1769061 MP MT] [instance_base.py:301]:
87++---------------------------------------------------------------------------------------+
88+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 60s Rev: 9bfa978 |
89++---------------------------------------------------------------------------------------+
90+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_3 OnGeirExecute RUNNING 10s |
91++---------------------------------------------------------------------------------------+
92+2026-09-28 02:15:44,378 [INFO] [1769082 auto_testcase_name_3 OnReturning] [profiling.py:63]:
93+########################
94+Performance result: Comparison result:
95+CST_PERF: None CST_GOLD: PASS,PASS
96+DYN_PERF: None DYN_GOLD:
97+PRECISION_STATUS: PASS PASSED: PASS
98+########################
99+
100+2026-09-28 02:15:48,904 [INFO] [1769082 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
101+2026-09-28 02:15:48,904 [INFO] [1769082 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_4.json
102+2026-09-28 02:15:48,905 [INFO] [1769082 auto_testcase_name_4 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
103+2026-09-28 02:15:49,876 [INFO] [1769061 MP MT] [instance_base.py:301]:
104++--------------------------------------------------------------------------------------------------------+
105+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 70s Rev: 9bfa978 |
106++--------------------------------------------------------------------------------------------------------+
107+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_4 OnGeirExecute RUNNING 0s |
108++--------------------------------------------------------------------------------------------------------+
109+2026-09-28 02:15:59,881 [INFO] [1769061 MP MT] [instance_base.py:301]:
110++---------------------------------------------------------------------------------------+
111+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 80s Rev: 9bfa978 |
112++---------------------------------------------------------------------------------------+
113+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_4 OnGeirExecute RUNNING 10s |
114++---------------------------------------------------------------------------------------+
115+2026-09-28 02:16:03,786 [INFO] [1769082 auto_testcase_name_4 OnReturning] [profiling.py:63]:
116+########################
117+Performance result: Comparison result:
118+CST_PERF: None CST_GOLD: PASS,PASS
119+DYN_PERF: None DYN_GOLD:
120+PRECISION_STATUS: PASS PASSED: PASS
121+########################
122+
123+2026-09-28 02:16:08,271 [INFO] [1769082 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
124+2026-09-28 02:16:08,271 [INFO] [1769082 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_5.json
125+2026-09-28 02:16:08,271 [INFO] [1769082 auto_testcase_name_5 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
126+2026-09-28 02:16:09,891 [INFO] [1769061 MP MT] [instance_base.py:301]:
127++--------------------------------------------------------------------------------------------------------+
128+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 90s Rev: 9bfa978 |
129++--------------------------------------------------------------------------------------------------------+
130+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_5 OnGeirExecute RUNNING 1s |
131++--------------------------------------------------------------------------------------------------------+
132+2026-09-28 02:16:19,897 [INFO] [1769061 MP MT] [instance_base.py:301]:
133++----------------------------------------------------------------------------------------------------------+
134+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 100s Rev: 9bfa978 |
135++----------------------------------------------------------------------------------------------------------+
136+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_5 OnGeirExecute RUNNING 11s |
137++----------------------------------------------------------------------------------------------------------+
138+2026-09-28 02:16:23,026 [INFO] [1769082 auto_testcase_name_5 OnReturning] [profiling.py:63]:
139+########################
140+Performance result: Comparison result:
141+CST_PERF: None CST_GOLD: PASS,PASS
142+DYN_PERF: None DYN_GOLD:
143+PRECISION_STATUS: PASS PASSED: PASS
144+########################
145+
146+2026-09-28 02:16:28,058 [INFO] [1769082 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
147+2026-09-28 02:16:28,059 [INFO] [1769082 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_6.json
148+2026-09-28 02:16:28,059 [INFO] [1769082 auto_testcase_name_6 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
149+2026-09-28 02:16:29,899 [INFO] [1769061 MP MT] [instance_base.py:301]:
150++---------------------------------------------------------------------------------------------------------+
151+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 110s Rev: 9bfa978 |
152++---------------------------------------------------------------------------------------------------------+
153+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_6 OnGeirExecute RUNNING 1s |
154++---------------------------------------------------------------------------------------------------------+
155+2026-09-28 02:16:39,906 [INFO] [1769061 MP MT] [instance_base.py:301]:
156++----------------------------------------------------------------------------------------------------------+
157+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 120s Rev: 9bfa978 |
158++----------------------------------------------------------------------------------------------------------+
159+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_6 OnGeirExecute RUNNING 11s |
160++----------------------------------------------------------------------------------------------------------+
161+2026-09-28 02:16:42,633 [INFO] [1769082 auto_testcase_name_6 OnReturning] [profiling.py:63]:
162+########################
163+Performance result: Comparison result:
164+CST_PERF: None CST_GOLD: PASS,PASS
165+DYN_PERF: None DYN_GOLD:
166+PRECISION_STATUS: PASS PASSED: PASS
167+########################
168+
169+2026-09-28 02:16:47,179 [INFO] [1769082 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
170+2026-09-28 02:16:47,180 [INFO] [1769082 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_7.json
171+2026-09-28 02:16:47,180 [INFO] [1769082 auto_testcase_name_7 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
172+2026-09-28 02:16:49,906 [INFO] [1769061 MP MT] [instance_base.py:301]:
173++---------------------------------------------------------------------------------------------------------+
174+| Version: 3.0.0 Summary (Device Total: 1) Progress: 60% 6 / 10 ET: 130s Rev: 9bfa978 |
175++---------------------------------------------------------------------------------------------------------+
176+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_7 OnGeirExecute RUNNING 2s |
177++---------------------------------------------------------------------------------------------------------+
178+2026-09-28 02:16:59,913 [INFO] [1769061 MP MT] [instance_base.py:301]:
179++----------------------------------------------------------------------------------------------------------+
180+| Version: 3.0.0 Summary (Device Total: 1) Progress: 60% 6 / 10 ET: 140s Rev: 9bfa978 |
181++----------------------------------------------------------------------------------------------------------+
182+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_7 OnGeirExecute RUNNING 12s |
183++----------------------------------------------------------------------------------------------------------+
184+2026-09-28 02:17:01,559 [INFO] [1769082 auto_testcase_name_7 OnReturning] [profiling.py:63]:
185+########################
186+Performance result: Comparison result:
187+CST_PERF: None CST_GOLD: PASS,PASS
188+DYN_PERF: None DYN_GOLD:
189+PRECISION_STATUS: PASS PASSED: PASS
190+########################
191+
192+2026-09-28 02:17:06,012 [INFO] [1769082 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
193+2026-09-28 02:17:06,013 [INFO] [1769082 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_8.json
194+2026-09-28 02:17:06,013 [INFO] [1769082 auto_testcase_name_8 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
195+2026-09-28 02:17:09,921 [INFO] [1769061 MP MT] [instance_base.py:301]:
196++---------------------------------------------------------------------------------------------------------+
197+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 150s Rev: 9bfa978 |
198++---------------------------------------------------------------------------------------------------------+
199+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_8 OnGeirExecute RUNNING 3s |
200++---------------------------------------------------------------------------------------------------------+
201+2026-09-28 02:17:19,923 [INFO] [1769061 MP MT] [instance_base.py:301]:
202++----------------------------------------------------------------------------------------------------------+
203+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 160s Rev: 9bfa978 |
204++----------------------------------------------------------------------------------------------------------+
205+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_8 OnGeirExecute RUNNING 13s |
206++----------------------------------------------------------------------------------------------------------+
207+2026-09-28 02:17:20,585 [INFO] [1769082 auto_testcase_name_8 OnReturning] [profiling.py:63]:
208+########################
209+Performance result: Comparison result:
210+CST_PERF: None CST_GOLD: PASS,PASS
211+DYN_PERF: None DYN_GOLD:
212+PRECISION_STATUS: PASS PASSED: PASS
213+########################
214+
215+2026-09-28 02:17:25,086 [INFO] [1769082 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
216+2026-09-28 02:17:25,086 [INFO] [1769082 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_9.json
217+2026-09-28 02:17:25,086 [INFO] [1769082 auto_testcase_name_9 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
218+2026-09-28 02:17:29,931 [INFO] [1769061 MP MT] [instance_base.py:301]:
219++---------------------------------------------------------------------------------------------------------+
220+| Version: 3.0.0 Summary (Device Total: 1) Progress: 80% 8 / 10 ET: 170s Rev: 9bfa978 |
221++---------------------------------------------------------------------------------------------------------+
222+| NPU:0 Ascend950PR_9579 | 1769082 _testcase_name_9 OnGeirExecute RUNNING 4s |
223++---------------------------------------------------------------------------------------------------------+
224+2026-09-28 02:17:39,584 [INFO] [1769082 auto_testcase_name_9 OnReturning] [profiling.py:63]:
225+########################
226+Performance result: Comparison result:
227+CST_PERF: None CST_GOLD: PASS,PASS
228+DYN_PERF: None DYN_GOLD:
229+PRECISION_STATUS: PASS PASSED: PASS
230+########################
231+
232+2026-09-28 02:17:39,932 [INFO] [1769061 MP MT] [instance_base.py:301]:
233++---------------------------------------------------------------------------------------------------------+
234+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 180s Rev: 9bfa978 |
235++---------------------------------------------------------------------------------------------------------+
236+| NPU:0 Ascend950PR_9579 | 1769082 testcase_name_10 OnXpuProfiling RUNNING 0s |
237++---------------------------------------------------------------------------------------------------------+
238+2026-09-28 02:17:44,087 [INFO] [1769082 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
239+2026-09-28 02:17:44,088 [INFO] [1769082 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_10.json
240+2026-09-28 02:17:44,088 [INFO] [1769082 auto_testcase_name_10 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
241+2026-09-28 02:17:49,938 [INFO] [1769061 MP MT] [instance_base.py:301]:
242++---------------------------------------------------------------------------------------------------------+
243+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 190s Rev: 9bfa978 |
244++---------------------------------------------------------------------------------------------------------+
245+| NPU:0 Ascend950PR_9579 | 1769082 testcase_name_10 OnGeirExecute RUNNING 5s |
246++---------------------------------------------------------------------------------------------------------+
247+2026-09-28 02:17:58,925 [INFO] [1769082 auto_testcase_name_10 OnReturning] [profiling.py:63]:
248+########################
249+Performance result: Comparison result:
250+CST_PERF: None CST_GOLD: PASS,PASS
251+DYN_PERF: None DYN_GOLD:
252+PRECISION_STATUS: PASS PASSED: PASS
253+########################
254+
255+2026-09-28 02:17:58,938 [INFO] [1769061 MP MT] [instance_base.py:147]: ttk Profiling complete
256+2026-09-28 02:17:58,964 [INFO] [1769061 MP MT] [heartbeat_manager.py:158]: tenant 4c9186aa8151 deleted on 127.0.0.1:19098
257+2026-09-28 02:17:58,973 [INFO] [1769061 MP MT] [instance_base.py:325]:
258+==================== TTK Test Summary ===================
259+ Total : 10
260+ PASS : 10
261+ FAIL : 0
262+ SKIP/NA : 0
263+ PassRate : 100.00% (10/10)
264+=========================================================
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_03.csv+11-0
@@ -0,0 +1,11 @@
1+testcase_name,network_name,op_name,input_shapes,input_dtypes,input_formats,output_shapes,output_dtypes,output_formats,input_ori_shapes,input_ori_formats,output_ori_shapes,output_ori_formats,attributes,input_data_ranges,precision_tolerances,absolute_precision,output_inplace_indexes,output_shape_unknown_indexes,is_enabled,remark,soc_series,priority,dump_file_prefix,manual_input_binaries,manual_golden_binaries
2+in_training_update_grad_gamma_beta_gen0611_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[2, 2, 7, 16], [2, 2, 7, 16]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 2, 7, 16], [1, 2, 7, 16]]","['float32', 'float32']","['NHWC', 'NHWC']","[[2, 2, 7, 16], [2, 2, 7, 16]]","['NHWC', 'NHWC']","[[1, 2, 7, 16], [1, 2, 7, 16]]","['NHWC', 'NHWC']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
3+in_training_update_grad_gamma_beta_gen0631_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[63, 7, 7, 1], [63, 7, 7, 1]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 7, 7, 1], [1, 7, 7, 1]]","['float32', 'float32']","['NHWC', 'NHWC']","[[63, 7, 7, 1], [63, 7, 7, 1]]","['NHWC', 'NHWC']","[[1, 7, 7, 1], [1, 7, 7, 1]]","['NHWC', 'NHWC']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
4+in_training_update_grad_gamma_beta_gen0652_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[127, 8, 7, 8], [127, 8, 7, 8]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 8, 7, 8], [1, 8, 7, 8]]","['float32', 'float32']","['NCHW', 'NCHW']","[[127, 8, 7, 8], [127, 8, 7, 8]]","['NCHW', 'NCHW']","[[1, 8, 7, 8], [1, 8, 7, 8]]","['NCHW', 'NCHW']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
5+in_training_update_grad_gamma_beta_gen0672_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[1024, 2, 7, 127], [1024, 2, 7, 127]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 2, 7, 127], [1, 2, 7, 127]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1024, 2, 7, 127], [1024, 2, 7, 127]]","['NCHW', 'NCHW']","[[1, 2, 7, 127], [1, 2, 7, 127]]","['NCHW', 'NCHW']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
6+in_training_update_grad_gamma_beta_gen0692_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[3, 4, 16, 4], [3, 4, 16, 4]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 4, 16, 4], [1, 4, 16, 4]]","['float32', 'float32']","['NCHW', 'NCHW']","[[3, 4, 16, 4], [3, 4, 16, 4]]","['NCHW', 'NCHW']","[[1, 4, 16, 4], [1, 4, 16, 4]]","['NCHW', 'NCHW']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
7+in_training_update_grad_gamma_beta_gen0714_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[1024, 2, 4, 32], [1024, 2, 4, 32]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 2, 4, 32], [1, 2, 4, 32]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1024, 2, 4, 32], [1024, 2, 4, 32]]","['NCHW', 'NCHW']","[[1, 2, 4, 32], [1, 2, 4, 32]]","['NCHW', 'NCHW']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
8+in_training_update_grad_gamma_beta_gen0734_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[31, 8, 2, 64], [31, 8, 2, 64]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 8, 2, 64], [1, 8, 2, 64]]","['float32', 'float32']","['NCHW', 'NCHW']","[[31, 8, 2, 64], [31, 8, 2, 64]]","['NCHW', 'NCHW']","[[1, 8, 2, 64], [1, 8, 2, 64]]","['NCHW', 'NCHW']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
9+in_training_update_grad_gamma_beta_gen0754_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[1024, 2, 15, 8], [1024, 2, 15, 8]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 2, 15, 8], [1, 2, 15, 8]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1024, 2, 15, 8], [1024, 2, 15, 8]]","['NCHW', 'NCHW']","[[1, 2, 15, 8], [1, 2, 15, 8]]","['NCHW', 'NCHW']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
10+in_training_update_grad_gamma_beta_gen0775_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[16, 16, 31, 31], [16, 16, 31, 31]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 16, 31, 31], [1, 16, 31, 31]]","['float32', 'float32']","['NHWC', 'NHWC']","[[16, 16, 31, 31], [16, 16, 31, 31]]","['NHWC', 'NHWC']","[[1, 16, 31, 31], [1, 16, 31, 31]]","['NHWC', 'NHWC']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
11+in_training_update_grad_gamma_beta_gen0795_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[8, 7, 7, 64], [8, 7, 7, 64]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 7, 7, 64], [1, 7, 7, 64]]","['float32', 'float32']","['NHWC', 'NHWC']","[[8, 7, 7, 64], [8, 7, 7, 64]]","['NHWC', 'NHWC']","[[1, 7, 7, 64], [1, 7, 7, 64]]","['NHWC', 'NHWC']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_03_run.log+270-0
@@ -0,0 +1,270 @@
1+2026-09-28 02:20:19,628 [DEBUG] [1781267 MainProcess MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
2+2026-09-28 02:20:19,630 [INFO] [1781267 MP MT] [bridge.py:441]: Command: ttk geir -i /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_03.csv
3+2026-09-28 02:20:19,631 [INFO] [1781267 MP MT] [instance_base.py:118]: TTK Profiling Start!
4+2026-09-28 02:20:19,631 [INFO] [1781267 MP MT] [instance_base.py:119]: Mode: ASCEND_ONBOARD
5+2026-09-28 02:20:19,635 [INFO] [1781267 MP MT] [instance_base.py:122]: Device Platform: Ascend950PR_9579
6+2026-09-28 02:20:19,635 [INFO] [1781267 MP MT] [instance_base.py:129]: Device Count: 1
7+2026-09-28 02:20:19,635 [INFO] [1781267 MP MT] [instance_base.py:372]: Preparing testcases ...
8+2026-09-28 02:20:19,635 [INFO] [1781267 MP MT] [instance_base.py:374]: Reading testcase file /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_03.csv ...
9+2026-09-28 02:20:19,635 [INFO] [1781267 MP MT] [instance_base.py:447]: Input testcase file is not a valid zip file, switch to normal csv mode...
10+2026-09-28 02:20:19,635 [INFO] [1781267 MP MT] [instance_base.py:723]: Reading testcases from /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_03.csv ...
11+2026-09-28 02:20:19,635 [WARNING] [1781267 TestcaseManager HeaderCheckTestcaseName] [testcase_manager.py:258]: Testcase name not found! It is important to add a testcase_name in order to identify your testcases
12+2026-09-28 02:20:19,635 [WARNING] [1781267 TestcaseManager HeaderCheckUnidentifiedHeaders] [testcase_manager.py:289]: Detected unidentified headers and will be ignored: ['\ufefftestcase_name']
13+2026-09-28 02:20:19,639 [INFO] [1781267 TestcaseManager auto_testcase_name_1] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
14+2026-09-28 02:20:19,692 [INFO] [1781267 MP MT] [instance_base.py:378]: Checking testcase name...
15+2026-09-28 02:20:19,692 [INFO] [1781267 MP MT] [instance_base.py:381]: Case number: 10
16+2026-09-28 02:20:19,704 [INFO] [1781267 MP MT] [heartbeat_manager.py:86]: Heartbeat subprocess started: pid=1781277, health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_939189fea3d3.json
17+2026-09-28 02:20:19,706 [INFO] [1781277 HB MT] [heartbeat.py:50]: HB started: tenant=939189fea3d3 health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_939189fea3d3.json n_endpoints=1 tls=False
18+2026-09-28 02:20:19,743 [INFO] [1781267 MP MT] [instance_base.py:463]: Initialize output csv file...
19+2026-09-28 02:20:19,788 [INFO] [1781267 MP MT] [instance_base.py:167]: Process per device: 1
20+2026-09-28 02:20:19,788 [INFO] [1781267 MP MT] [instance_base.py:169]: Preparing Task Executors...
21+2026-09-28 02:20:19,884 [DEBUG] [1781289 D0P0 MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
22+2026-09-28 02:20:19,892 [INFO] [1781267 MP MT] [instance_base.py:301]:
23++--------------------------------------------------------------------------------------+
24+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 0s Rev: 9bfa978 |
25++--------------------------------------------------------------------------------------+
26+| NPU:0 Ascend950PR_9579 | 1781289 D0P0 UNKNOWN IDLE 0s |
27++--------------------------------------------------------------------------------------+
28+2026-09-28 02:20:19,930 [INFO] [1781289 auto_testcase_name_1 OnGenerateInput] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
29+2026-09-28 02:20:22,671 [WARNING] [1781289 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:550]: 424 retry (1/5): {"error": "missing spec dependency: golden", "missing": "golden"}
30+2026-09-28 02:20:22,751 [INFO] [1781289 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:357]: 424: synced 'golden.py' for missing module 'golden'
31+2026-09-28 02:20:27,230 [INFO] [1781289 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
32+2026-09-28 02:20:27,230 [INFO] [1781289 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_1.json
33+2026-09-28 02:20:27,230 [INFO] [1781289 auto_testcase_name_1 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
34+2026-09-28 02:20:29,899 [INFO] [1781267 MP MT] [instance_base.py:301]:
35++--------------------------------------------------------------------------------------+
36+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 10s Rev: 9bfa978 |
37++--------------------------------------------------------------------------------------+
38+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_1 OnGeirExecute RUNNING 2s |
39++--------------------------------------------------------------------------------------+
40+2026-09-28 02:20:39,907 [INFO] [1781267 MP MT] [instance_base.py:301]:
41++---------------------------------------------------------------------------------------+
42+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 20s Rev: 9bfa978 |
43++---------------------------------------------------------------------------------------+
44+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_1 OnGeirExecute RUNNING 12s |
45++---------------------------------------------------------------------------------------+
46+2026-09-28 02:20:42,750 [INFO] [1781289 auto_testcase_name_1 OnReturning] [profiling.py:63]:
47+########################
48+Performance result: Comparison result:
49+CST_PERF: None CST_GOLD: PASS,PASS
50+DYN_PERF: None DYN_GOLD:
51+PRECISION_STATUS: PASS PASSED: PASS
52+########################
53+
54+2026-09-28 02:20:47,291 [INFO] [1781289 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
55+2026-09-28 02:20:47,291 [INFO] [1781289 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_2.json
56+2026-09-28 02:20:47,292 [INFO] [1781289 auto_testcase_name_2 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
57+2026-09-28 02:20:49,909 [INFO] [1781267 MP MT] [instance_base.py:301]:
58++--------------------------------------------------------------------------------------------------------+
59+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 30s Rev: 9bfa978 |
60++--------------------------------------------------------------------------------------------------------+
61+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_2 OnGeirExecute RUNNING 2s |
62++--------------------------------------------------------------------------------------------------------+
63+2026-09-28 02:20:59,914 [INFO] [1781267 MP MT] [instance_base.py:301]:
64++---------------------------------------------------------------------------------------+
65+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 40s Rev: 9bfa978 |
66++---------------------------------------------------------------------------------------+
67+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_2 OnGeirExecute RUNNING 12s |
68++---------------------------------------------------------------------------------------+
69+2026-09-28 02:21:02,794 [INFO] [1781289 auto_testcase_name_2 OnReturning] [profiling.py:63]:
70+########################
71+Performance result: Comparison result:
72+CST_PERF: None CST_GOLD: PASS,PASS
73+DYN_PERF: None DYN_GOLD:
74+PRECISION_STATUS: PASS PASSED: PASS
75+########################
76+
77+2026-09-28 02:21:07,295 [INFO] [1781289 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
78+2026-09-28 02:21:07,296 [INFO] [1781289 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_3.json
79+2026-09-28 02:21:07,296 [INFO] [1781289 auto_testcase_name_3 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
80+2026-09-28 02:21:09,916 [INFO] [1781267 MP MT] [instance_base.py:301]:
81++--------------------------------------------------------------------------------------------------------+
82+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 50s Rev: 9bfa978 |
83++--------------------------------------------------------------------------------------------------------+
84+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_3 OnGeirExecute RUNNING 2s |
85++--------------------------------------------------------------------------------------------------------+
86+2026-09-28 02:21:19,920 [INFO] [1781267 MP MT] [instance_base.py:301]:
87++---------------------------------------------------------------------------------------+
88+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 60s Rev: 9bfa978 |
89++---------------------------------------------------------------------------------------+
90+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_3 OnGeirExecute RUNNING 12s |
91++---------------------------------------------------------------------------------------+
92+2026-09-28 02:21:22,466 [INFO] [1781289 auto_testcase_name_3 OnReturning] [profiling.py:63]:
93+########################
94+Performance result: Comparison result:
95+CST_PERF: None CST_GOLD: PASS,PASS
96+DYN_PERF: None DYN_GOLD:
97+PRECISION_STATUS: PASS PASSED: PASS
98+########################
99+
100+2026-09-28 02:21:27,553 [INFO] [1781289 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
101+2026-09-28 02:21:27,553 [INFO] [1781289 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_4.json
102+2026-09-28 02:21:27,553 [INFO] [1781289 auto_testcase_name_4 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
103+2026-09-28 02:21:29,929 [INFO] [1781267 MP MT] [instance_base.py:301]:
104++--------------------------------------------------------------------------------------------------------+
105+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 70s Rev: 9bfa978 |
106++--------------------------------------------------------------------------------------------------------+
107+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_4 OnGeirExecute RUNNING 2s |
108++--------------------------------------------------------------------------------------------------------+
109+2026-09-28 02:21:39,932 [INFO] [1781267 MP MT] [instance_base.py:301]:
110++---------------------------------------------------------------------------------------+
111+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 80s Rev: 9bfa978 |
112++---------------------------------------------------------------------------------------+
113+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_4 OnGeirExecute RUNNING 12s |
114++---------------------------------------------------------------------------------------+
115+2026-09-28 02:21:42,803 [INFO] [1781289 auto_testcase_name_4 OnReturning] [profiling.py:63]:
116+########################
117+Performance result: Comparison result:
118+CST_PERF: None CST_GOLD: PASS,PASS
119+DYN_PERF: None DYN_GOLD:
120+PRECISION_STATUS: PASS PASSED: PASS
121+########################
122+
123+2026-09-28 02:21:47,295 [INFO] [1781289 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
124+2026-09-28 02:21:47,296 [INFO] [1781289 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_5.json
125+2026-09-28 02:21:47,296 [INFO] [1781289 auto_testcase_name_5 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
126+2026-09-28 02:21:49,935 [INFO] [1781267 MP MT] [instance_base.py:301]:
127++--------------------------------------------------------------------------------------------------------+
128+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 90s Rev: 9bfa978 |
129++--------------------------------------------------------------------------------------------------------+
130+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_5 OnGeirExecute RUNNING 2s |
131++--------------------------------------------------------------------------------------------------------+
132+2026-09-28 02:21:59,940 [INFO] [1781267 MP MT] [instance_base.py:301]:
133++----------------------------------------------------------------------------------------------------------+
134+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 100s Rev: 9bfa978 |
135++----------------------------------------------------------------------------------------------------------+
136+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_5 OnGeirExecute RUNNING 12s |
137++----------------------------------------------------------------------------------------------------------+
138+2026-09-28 02:22:03,519 [INFO] [1781289 auto_testcase_name_5 OnReturning] [profiling.py:63]:
139+########################
140+Performance result: Comparison result:
141+CST_PERF: None CST_GOLD: PASS,PASS
142+DYN_PERF: None DYN_GOLD:
143+PRECISION_STATUS: PASS PASSED: PASS
144+########################
145+
146+2026-09-28 02:22:08,003 [INFO] [1781289 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
147+2026-09-28 02:22:08,003 [INFO] [1781289 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_6.json
148+2026-09-28 02:22:08,004 [INFO] [1781289 auto_testcase_name_6 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
149+2026-09-28 02:22:09,941 [INFO] [1781267 MP MT] [instance_base.py:301]:
150++---------------------------------------------------------------------------------------------------------+
151+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 110s Rev: 9bfa978 |
152++---------------------------------------------------------------------------------------------------------+
153+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_6 OnGeirExecute RUNNING 1s |
154++---------------------------------------------------------------------------------------------------------+
155+2026-09-28 02:22:19,949 [INFO] [1781267 MP MT] [instance_base.py:301]:
156++----------------------------------------------------------------------------------------------------------+
157+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 120s Rev: 9bfa978 |
158++----------------------------------------------------------------------------------------------------------+
159+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_6 OnGeirExecute RUNNING 11s |
160++----------------------------------------------------------------------------------------------------------+
161+2026-09-28 02:22:22,937 [INFO] [1781289 auto_testcase_name_6 OnReturning] [profiling.py:63]:
162+########################
163+Performance result: Comparison result:
164+CST_PERF: None CST_GOLD: PASS,PASS
165+DYN_PERF: None DYN_GOLD:
166+PRECISION_STATUS: PASS PASSED: PASS
167+########################
168+
169+2026-09-28 02:22:27,453 [INFO] [1781289 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
170+2026-09-28 02:22:27,454 [INFO] [1781289 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_7.json
171+2026-09-28 02:22:27,454 [INFO] [1781289 auto_testcase_name_7 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
172+2026-09-28 02:22:29,956 [INFO] [1781267 MP MT] [instance_base.py:301]:
173++---------------------------------------------------------------------------------------------------------+
174+| Version: 3.0.0 Summary (Device Total: 1) Progress: 60% 6 / 10 ET: 130s Rev: 9bfa978 |
175++---------------------------------------------------------------------------------------------------------+
176+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_7 OnGeirExecute RUNNING 2s |
177++---------------------------------------------------------------------------------------------------------+
178+2026-09-28 02:22:39,962 [INFO] [1781267 MP MT] [instance_base.py:301]:
179++----------------------------------------------------------------------------------------------------------+
180+| Version: 3.0.0 Summary (Device Total: 1) Progress: 60% 6 / 10 ET: 140s Rev: 9bfa978 |
181++----------------------------------------------------------------------------------------------------------+
182+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_7 OnGeirExecute RUNNING 12s |
183++----------------------------------------------------------------------------------------------------------+
184+2026-09-28 02:22:43,467 [INFO] [1781289 auto_testcase_name_7 OnReturning] [profiling.py:63]:
185+########################
186+Performance result: Comparison result:
187+CST_PERF: None CST_GOLD: PASS,PASS
188+DYN_PERF: None DYN_GOLD:
189+PRECISION_STATUS: PASS PASSED: PASS
190+########################
191+
192+2026-09-28 02:22:48,028 [INFO] [1781289 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
193+2026-09-28 02:22:48,028 [INFO] [1781289 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_8.json
194+2026-09-28 02:22:48,029 [INFO] [1781289 auto_testcase_name_8 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
195+2026-09-28 02:22:49,963 [INFO] [1781267 MP MT] [instance_base.py:301]:
196++---------------------------------------------------------------------------------------------------------+
197+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 150s Rev: 9bfa978 |
198++---------------------------------------------------------------------------------------------------------+
199+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_8 OnGeirExecute RUNNING 1s |
200++---------------------------------------------------------------------------------------------------------+
201+2026-09-28 02:22:59,970 [INFO] [1781267 MP MT] [instance_base.py:301]:
202++----------------------------------------------------------------------------------------------------------+
203+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 160s Rev: 9bfa978 |
204++----------------------------------------------------------------------------------------------------------+
205+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_8 OnGeirExecute RUNNING 11s |
206++----------------------------------------------------------------------------------------------------------+
207+2026-09-28 02:23:03,244 [INFO] [1781289 auto_testcase_name_8 OnReturning] [profiling.py:63]:
208+########################
209+Performance result: Comparison result:
210+CST_PERF: None CST_GOLD: PASS,PASS
211+DYN_PERF: None DYN_GOLD:
212+PRECISION_STATUS: PASS PASSED: PASS
213+########################
214+
215+2026-09-28 02:23:08,601 [INFO] [1781289 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
216+2026-09-28 02:23:08,601 [INFO] [1781289 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_9.json
217+2026-09-28 02:23:08,602 [INFO] [1781289 auto_testcase_name_9 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
218+2026-09-28 02:23:09,971 [INFO] [1781267 MP MT] [instance_base.py:301]:
219++---------------------------------------------------------------------------------------------------------+
220+| Version: 3.0.0 Summary (Device Total: 1) Progress: 80% 8 / 10 ET: 170s Rev: 9bfa978 |
221++---------------------------------------------------------------------------------------------------------+
222+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_9 OnGeirExecute RUNNING 1s |
223++---------------------------------------------------------------------------------------------------------+
224+2026-09-28 02:23:19,976 [INFO] [1781267 MP MT] [instance_base.py:301]:
225++----------------------------------------------------------------------------------------------------------+
226+| Version: 3.0.0 Summary (Device Total: 1) Progress: 80% 8 / 10 ET: 180s Rev: 9bfa978 |
227++----------------------------------------------------------------------------------------------------------+
228+| NPU:0 Ascend950PR_9579 | 1781289 _testcase_name_9 OnGeirExecute RUNNING 11s |
229++----------------------------------------------------------------------------------------------------------+
230+2026-09-28 02:23:24,845 [INFO] [1781289 auto_testcase_name_9 OnReturning] [profiling.py:63]:
231+########################
232+Performance result: Comparison result:
233+CST_PERF: None CST_GOLD: PASS,PASS
234+DYN_PERF: None DYN_GOLD:
235+PRECISION_STATUS: PASS PASSED: PASS
236+########################
237+
238+2026-09-28 02:23:29,340 [INFO] [1781289 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
239+2026-09-28 02:23:29,340 [INFO] [1781289 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_10.json
240+2026-09-28 02:23:29,340 [INFO] [1781289 auto_testcase_name_10 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
241+2026-09-28 02:23:29,985 [INFO] [1781267 MP MT] [instance_base.py:301]:
242++---------------------------------------------------------------------------------------------------------+
243+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 190s Rev: 9bfa978 |
244++---------------------------------------------------------------------------------------------------------+
245+| NPU:0 Ascend950PR_9579 | 1781289 testcase_name_10 OnGeirExecute RUNNING 0s |
246++---------------------------------------------------------------------------------------------------------+
247+2026-09-28 02:23:39,991 [INFO] [1781267 MP MT] [instance_base.py:301]:
248++----------------------------------------------------------------------------------------------------------+
249+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 200s Rev: 9bfa978 |
250++----------------------------------------------------------------------------------------------------------+
251+| NPU:0 Ascend950PR_9579 | 1781289 testcase_name_10 OnGeirExecute RUNNING 10s |
252++----------------------------------------------------------------------------------------------------------+
253+2026-09-28 02:23:44,934 [INFO] [1781289 auto_testcase_name_10 OnReturning] [profiling.py:63]:
254+########################
255+Performance result: Comparison result:
256+CST_PERF: None CST_GOLD: PASS,PASS
257+DYN_PERF: None DYN_GOLD:
258+PRECISION_STATUS: PASS PASSED: PASS
259+########################
260+
261+2026-09-28 02:23:44,944 [INFO] [1781267 MP MT] [instance_base.py:147]: ttk Profiling complete
262+2026-09-28 02:23:44,968 [INFO] [1781267 MP MT] [heartbeat_manager.py:158]: tenant 939189fea3d3 deleted on 127.0.0.1:19098
263+2026-09-28 02:23:44,979 [INFO] [1781267 MP MT] [instance_base.py:325]:
264+==================== TTK Test Summary ===================
265+ Total : 10
266+ PASS : 10
267+ FAIL : 0
268+ SKIP/NA : 0
269+ PassRate : 100.00% (10/10)
270+=========================================================
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_04.csv+11-0
@@ -0,0 +1,11 @@
1+testcase_name,network_name,op_name,input_shapes,input_dtypes,input_formats,output_shapes,output_dtypes,output_formats,input_ori_shapes,input_ori_formats,output_ori_shapes,output_ori_formats,attributes,input_data_ranges,precision_tolerances,absolute_precision,output_inplace_indexes,output_shape_unknown_indexes,is_enabled,remark,soc_series,priority,dump_file_prefix,manual_input_binaries,manual_golden_binaries
2+in_training_update_grad_gamma_beta_gen0815_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[4, 2, 7, 15], [4, 2, 7, 15]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 2, 7, 15], [1, 2, 7, 15]]","['float32', 'float32']","['NHWC', 'NHWC']","[[4, 2, 7, 15], [4, 2, 7, 15]]","['NHWC', 'NHWC']","[[1, 2, 7, 15], [1, 2, 7, 15]]","['NHWC', 'NHWC']",{},"((-1, 1), (-1, 1))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
3+in_training_update_grad_gamma_beta_gen0836_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[8, 8, 15, 647], [8, 8, 15, 647]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 8, 15, 647], [1, 8, 15, 647]]","['float32', 'float32']","['NCHW', 'NCHW']","[[8, 8, 15, 647], [8, 8, 15, 647]]","['NCHW', 'NCHW']","[[1, 8, 15, 647], [1, 8, 15, 647]]","['NCHW', 'NCHW']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
4+in_training_update_grad_gamma_beta_gen0856_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[1024, 4, 5, 5], [1024, 4, 5, 5]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 4, 5, 5], [1, 4, 5, 5]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1024, 4, 5, 5], [1024, 4, 5, 5]]","['NCHW', 'NCHW']","[[1, 4, 5, 5], [1, 4, 5, 5]]","['NCHW', 'NCHW']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
5+in_training_update_grad_gamma_beta_gen0876_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[257, 7, 8, 1], [257, 7, 8, 1]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 7, 8, 1], [1, 7, 8, 1]]","['float32', 'float32']","['NCHW', 'NCHW']","[[257, 7, 8, 1], [257, 7, 8, 1]]","['NCHW', 'NCHW']","[[1, 7, 8, 1], [1, 7, 8, 1]]","['NCHW', 'NCHW']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
6+in_training_update_grad_gamma_beta_gen0896_nchw_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[257, 31, 1, 647], [257, 31, 1, 647]]","['float32', 'float32']","['NCHW', 'NCHW']","[[1, 31, 1, 647], [1, 31, 1, 647]]","['float32', 'float32']","['NCHW', 'NCHW']","[[257, 31, 1, 647], [257, 31, 1, 647]]","['NCHW', 'NCHW']","[[1, 31, 1, 647], [1, 31, 1, 647]]","['NCHW', 'NCHW']",{},"((-100, 100), (-100, 100))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NCHW,,,,(),()
7+in_training_update_grad_gamma_beta_gen0917_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[3, 4, 8, 649], [3, 4, 8, 649]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 4, 8, 649], [1, 4, 8, 649]]","['float32', 'float32']","['NHWC', 'NHWC']","[[3, 4, 8, 649], [3, 4, 8, 649]]","['NHWC', 'NHWC']","[[1, 4, 8, 649], [1, 4, 8, 649]]","['NHWC', 'NHWC']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
8+in_training_update_grad_gamma_beta_gen0937_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[63, 5, 31, 257], [63, 5, 31, 257]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 5, 31, 257], [1, 5, 31, 257]]","['float32', 'float32']","['NHWC', 'NHWC']","[[63, 5, 31, 257], [63, 5, 31, 257]]","['NHWC', 'NHWC']","[[1, 5, 31, 257], [1, 5, 31, 257]]","['NHWC', 'NHWC']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
9+in_training_update_grad_gamma_beta_gen0957_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[32, 16, 2, 31], [32, 16, 2, 31]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 16, 2, 31], [1, 16, 2, 31]]","['float32', 'float32']","['NHWC', 'NHWC']","[[32, 16, 2, 31], [32, 16, 2, 31]]","['NHWC', 'NHWC']","[[1, 16, 2, 31], [1, 16, 2, 31]]","['NHWC', 'NHWC']",{},"((-0.001, 0.001), (-0.001, 0.001))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
10+in_training_update_grad_gamma_beta_gen0979_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[1024, 2, 8, 7], [1024, 2, 8, 7]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 2, 8, 7], [1, 2, 8, 7]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1024, 2, 8, 7], [1024, 2, 8, 7]]","['NHWC', 'NHWC']","[[1, 2, 8, 7], [1, 2, 8, 7]]","['NHWC', 'NHWC']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
11+in_training_update_grad_gamma_beta_gen0999_nhwc_f32,UNKNOWN,in_training_update_grad_gamma_beta,"[[1, 5, 7, 7], [1, 5, 7, 7]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 5, 7, 7], [1, 5, 7, 7]]","['float32', 'float32']","['NHWC', 'NHWC']","[[1, 5, 7, 7], [1, 5, 7, 7]]","['NHWC', 'NHWC']","[[1, 5, 7, 7], [1, 5, 7, 7]]","['NHWC', 'NHWC']",{},"((-1e-30, 1e-30), (-1e-30, 1e-30))","((0.0001, 0.0001), (0.0001, 0.0001))","(0.0001, 0.0001)",(),(),1,tiling_key=0; public NHWC,,,,(),()
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/chunks/chunk_04_run.log+276-0
@@ -0,0 +1,276 @@
1+2026-09-28 02:25:59,646 [DEBUG] [1794572 MainProcess MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
2+2026-09-28 02:25:59,647 [INFO] [1794572 MP MT] [bridge.py:441]: Command: ttk geir -i /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_04.csv
3+2026-09-28 02:25:59,649 [INFO] [1794572 MP MT] [instance_base.py:118]: TTK Profiling Start!
4+2026-09-28 02:25:59,649 [INFO] [1794572 MP MT] [instance_base.py:119]: Mode: ASCEND_ONBOARD
5+2026-09-28 02:25:59,652 [INFO] [1794572 MP MT] [instance_base.py:122]: Device Platform: Ascend950PR_9579
6+2026-09-28 02:25:59,653 [INFO] [1794572 MP MT] [instance_base.py:129]: Device Count: 1
7+2026-09-28 02:25:59,653 [INFO] [1794572 MP MT] [instance_base.py:372]: Preparing testcases ...
8+2026-09-28 02:25:59,653 [INFO] [1794572 MP MT] [instance_base.py:374]: Reading testcase file /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_04.csv ...
9+2026-09-28 02:25:59,653 [INFO] [1794572 MP MT] [instance_base.py:447]: Input testcase file is not a valid zip file, switch to normal csv mode...
10+2026-09-28 02:25:59,653 [INFO] [1794572 MP MT] [instance_base.py:723]: Reading testcases from /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/quality_hardening/geir_crosscheck_20260928/chunks/chunk_04.csv ...
11+2026-09-28 02:25:59,653 [WARNING] [1794572 TestcaseManager HeaderCheckTestcaseName] [testcase_manager.py:258]: Testcase name not found! It is important to add a testcase_name in order to identify your testcases
12+2026-09-28 02:25:59,653 [WARNING] [1794572 TestcaseManager HeaderCheckUnidentifiedHeaders] [testcase_manager.py:289]: Detected unidentified headers and will be ignored: ['\ufefftestcase_name']
13+2026-09-28 02:25:59,656 [INFO] [1794572 TestcaseManager auto_testcase_name_1] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
14+2026-09-28 02:25:59,709 [INFO] [1794572 MP MT] [instance_base.py:378]: Checking testcase name...
15+2026-09-28 02:25:59,709 [INFO] [1794572 MP MT] [instance_base.py:381]: Case number: 10
16+2026-09-28 02:25:59,721 [INFO] [1794572 MP MT] [heartbeat_manager.py:86]: Heartbeat subprocess started: pid=1794574, health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_c6a56ec8ab24.json
17+2026-09-28 02:25:59,723 [INFO] [1794574 HB MT] [heartbeat.py:50]: HB started: tenant=c6a56ec8ab24 health_path=/workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/.ttk/xpu_health_c6a56ec8ab24.json n_endpoints=1 tls=False
18+2026-09-28 02:25:59,757 [INFO] [1794572 MP MT] [instance_base.py:463]: Initialize output csv file...
19+2026-09-28 02:25:59,802 [INFO] [1794572 MP MT] [instance_base.py:167]: Process per device: 1
20+2026-09-28 02:25:59,802 [INFO] [1794572 MP MT] [instance_base.py:169]: Preparing Task Executors...
21+2026-09-28 02:25:59,900 [DEBUG] [1794586 D0P0 MainThread] [loader.py:157]: Loading config: /root/.config/ttk/ttk.conf.yaml
22+2026-09-28 02:25:59,907 [INFO] [1794572 MP MT] [instance_base.py:301]:
23++--------------------------------------------------------------------------------------+
24+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 0s Rev: 9bfa978 |
25++--------------------------------------------------------------------------------------+
26+| NPU:0 Ascend950PR_9579 | 1794586 D0P0 UNKNOWN IDLE 0s |
27++--------------------------------------------------------------------------------------+
28+2026-09-28 02:25:59,944 [INFO] [1794586 auto_testcase_name_1 OnGenerateInput] [op_info_keeper.py:353]: Loading vendor op info from: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/Ascend/cann-9.2.0/opp/vendors/custom_nn/op_impl/ai_core/tbe/config/ascend950/aic-ascend950-ops-info-nn.json
29+2026-09-28 02:26:02,640 [WARNING] [1794586 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:550]: 424 retry (1/5): {"error": "missing spec dependency: golden", "missing": "golden"}
30+2026-09-28 02:26:02,724 [INFO] [1794586 auto_testcase_name_1 OnXpuProfiling] [dispatcher.py:357]: 424: synced 'golden.py' for missing module 'golden'
31+2026-09-28 02:26:07,185 [INFO] [1794586 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
32+2026-09-28 02:26:07,185 [INFO] [1794586 auto_testcase_name_1 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_1.json
33+2026-09-28 02:26:07,186 [INFO] [1794586 auto_testcase_name_1 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
34+2026-09-28 02:26:09,906 [INFO] [1794572 MP MT] [instance_base.py:301]:
35++--------------------------------------------------------------------------------------+
36+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 10s Rev: 9bfa978 |
37++--------------------------------------------------------------------------------------+
38+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_1 OnGeirExecute RUNNING 2s |
39++--------------------------------------------------------------------------------------+
40+2026-09-28 02:26:19,914 [INFO] [1794572 MP MT] [instance_base.py:301]:
41++---------------------------------------------------------------------------------------+
42+| Version: 3.0.0 Summary (Device Total: 1) Progress: 0% 0 / 10 ET: 20s Rev: 9bfa978 |
43++---------------------------------------------------------------------------------------+
44+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_1 OnGeirExecute RUNNING 12s |
45++---------------------------------------------------------------------------------------+
46+2026-09-28 02:26:22,697 [INFO] [1794586 auto_testcase_name_1 OnReturning] [profiling.py:63]:
47+########################
48+Performance result: Comparison result:
49+CST_PERF: None CST_GOLD: PASS,PASS
50+DYN_PERF: None DYN_GOLD:
51+PRECISION_STATUS: PASS PASSED: PASS
52+########################
53+
54+2026-09-28 02:26:27,812 [INFO] [1794586 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
55+2026-09-28 02:26:27,812 [INFO] [1794586 auto_testcase_name_2 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_2.json
56+2026-09-28 02:26:27,812 [INFO] [1794586 auto_testcase_name_2 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
57+2026-09-28 02:26:29,916 [INFO] [1794572 MP MT] [instance_base.py:301]:
58++--------------------------------------------------------------------------------------------------------+
59+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 30s Rev: 9bfa978 |
60++--------------------------------------------------------------------------------------------------------+
61+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_2 OnGeirExecute RUNNING 2s |
62++--------------------------------------------------------------------------------------------------------+
63+2026-09-28 02:26:39,922 [INFO] [1794572 MP MT] [instance_base.py:301]:
64++---------------------------------------------------------------------------------------+
65+| Version: 3.0.0 Summary (Device Total: 1) Progress: 10% 1 / 10 ET: 40s Rev: 9bfa978 |
66++---------------------------------------------------------------------------------------+
67+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_2 OnGeirExecute RUNNING 12s |
68++---------------------------------------------------------------------------------------+
69+2026-09-28 02:26:42,845 [INFO] [1794586 auto_testcase_name_2 OnReturning] [profiling.py:63]:
70+########################
71+Performance result: Comparison result:
72+CST_PERF: None CST_GOLD: PASS,PASS
73+DYN_PERF: None DYN_GOLD:
74+PRECISION_STATUS: PASS PASSED: PASS
75+########################
76+
77+2026-09-28 02:26:47,438 [INFO] [1794586 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
78+2026-09-28 02:26:47,438 [INFO] [1794586 auto_testcase_name_3 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_3.json
79+2026-09-28 02:26:47,438 [INFO] [1794586 auto_testcase_name_3 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
80+2026-09-28 02:26:49,923 [INFO] [1794572 MP MT] [instance_base.py:301]:
81++--------------------------------------------------------------------------------------------------------+
82+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 50s Rev: 9bfa978 |
83++--------------------------------------------------------------------------------------------------------+
84+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_3 OnGeirExecute RUNNING 2s |
85++--------------------------------------------------------------------------------------------------------+
86+2026-09-28 02:26:59,929 [INFO] [1794572 MP MT] [instance_base.py:301]:
87++---------------------------------------------------------------------------------------+
88+| Version: 3.0.0 Summary (Device Total: 1) Progress: 20% 2 / 10 ET: 60s Rev: 9bfa978 |
89++---------------------------------------------------------------------------------------+
90+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_3 OnGeirExecute RUNNING 12s |
91++---------------------------------------------------------------------------------------+
92+2026-09-28 02:27:02,884 [INFO] [1794586 auto_testcase_name_3 OnReturning] [profiling.py:63]:
93+########################
94+Performance result: Comparison result:
95+CST_PERF: None CST_GOLD: PASS,PASS
96+DYN_PERF: None DYN_GOLD:
97+PRECISION_STATUS: PASS PASSED: PASS
98+########################
99+
100+2026-09-28 02:27:07,288 [INFO] [1794586 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
101+2026-09-28 02:27:07,288 [INFO] [1794586 auto_testcase_name_4 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_4.json
102+2026-09-28 02:27:07,289 [INFO] [1794586 auto_testcase_name_4 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
103+2026-09-28 02:27:09,936 [INFO] [1794572 MP MT] [instance_base.py:301]:
104++--------------------------------------------------------------------------------------------------------+
105+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 70s Rev: 9bfa978 |
106++--------------------------------------------------------------------------------------------------------+
107+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_4 OnGeirExecute RUNNING 2s |
108++--------------------------------------------------------------------------------------------------------+
109+2026-09-28 02:27:19,941 [INFO] [1794572 MP MT] [instance_base.py:301]:
110++---------------------------------------------------------------------------------------+
111+| Version: 3.0.0 Summary (Device Total: 1) Progress: 30% 3 / 10 ET: 80s Rev: 9bfa978 |
112++---------------------------------------------------------------------------------------+
113+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_4 OnGeirExecute RUNNING 12s |
114++---------------------------------------------------------------------------------------+
115+2026-09-28 02:27:23,154 [INFO] [1794586 auto_testcase_name_4 OnReturning] [profiling.py:63]:
116+########################
117+Performance result: Comparison result:
118+CST_PERF: None CST_GOLD: PASS,PASS
119+DYN_PERF: None DYN_GOLD:
120+PRECISION_STATUS: PASS PASSED: PASS
121+########################
122+
123+2026-09-28 02:27:29,943 [INFO] [1794572 MP MT] [instance_base.py:301]:
124++--------------------------------------------------------------------------------------------------------+
125+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 90s Rev: 9bfa978 |
126++--------------------------------------------------------------------------------------------------------+
127+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_5 OnXpuProfiling RUNNING 6s |
128++--------------------------------------------------------------------------------------------------------+
129+2026-09-28 02:27:38,100 [INFO] [1794586 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
130+2026-09-28 02:27:38,101 [INFO] [1794586 auto_testcase_name_5 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_5.json
131+2026-09-28 02:27:38,101 [INFO] [1794586 auto_testcase_name_5 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
132+2026-09-28 02:27:39,950 [INFO] [1794572 MP MT] [instance_base.py:301]:
133++---------------------------------------------------------------------------------------------------------+
134+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 100s Rev: 9bfa978 |
135++---------------------------------------------------------------------------------------------------------+
136+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_5 OnGeirExecute RUNNING 1s |
137++---------------------------------------------------------------------------------------------------------+
138+2026-09-28 02:27:49,957 [INFO] [1794572 MP MT] [instance_base.py:301]:
139++----------------------------------------------------------------------------------------------------------+
140+| Version: 3.0.0 Summary (Device Total: 1) Progress: 40% 4 / 10 ET: 110s Rev: 9bfa978 |
141++----------------------------------------------------------------------------------------------------------+
142+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_5 OnGeirExecute RUNNING 11s |
143++----------------------------------------------------------------------------------------------------------+
144+2026-09-28 02:27:52,947 [INFO] [1794586 auto_testcase_name_5 OnReturning] [profiling.py:63]:
145+########################
146+Performance result: Comparison result:
147+CST_PERF: None CST_GOLD: PASS,PASS
148+DYN_PERF: None DYN_GOLD:
149+PRECISION_STATUS: PASS PASSED: PASS
150+########################
151+
152+2026-09-28 02:27:57,577 [INFO] [1794586 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
153+2026-09-28 02:27:57,578 [INFO] [1794586 auto_testcase_name_6 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_6.json
154+2026-09-28 02:27:57,578 [INFO] [1794586 auto_testcase_name_6 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
155+2026-09-28 02:27:59,958 [INFO] [1794572 MP MT] [instance_base.py:301]:
156++---------------------------------------------------------------------------------------------------------+
157+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 120s Rev: 9bfa978 |
158++---------------------------------------------------------------------------------------------------------+
159+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_6 OnGeirExecute RUNNING 2s |
160++---------------------------------------------------------------------------------------------------------+
161+2026-09-28 02:28:09,964 [INFO] [1794572 MP MT] [instance_base.py:301]:
162++----------------------------------------------------------------------------------------------------------+
163+| Version: 3.0.0 Summary (Device Total: 1) Progress: 50% 5 / 10 ET: 130s Rev: 9bfa978 |
164++----------------------------------------------------------------------------------------------------------+
165+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_6 OnGeirExecute RUNNING 12s |
166++----------------------------------------------------------------------------------------------------------+
167+2026-09-28 02:28:12,625 [INFO] [1794586 auto_testcase_name_6 OnReturning] [profiling.py:63]:
168+########################
169+Performance result: Comparison result:
170+CST_PERF: None CST_GOLD: PASS,PASS
171+DYN_PERF: None DYN_GOLD:
172+PRECISION_STATUS: PASS PASSED: PASS
173+########################
174+
175+2026-09-28 02:28:18,342 [INFO] [1794586 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
176+2026-09-28 02:28:18,342 [INFO] [1794586 auto_testcase_name_7 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_7.json
177+2026-09-28 02:28:18,342 [INFO] [1794586 auto_testcase_name_7 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
178+2026-09-28 02:28:19,966 [INFO] [1794572 MP MT] [instance_base.py:301]:
179++---------------------------------------------------------------------------------------------------------+
180+| Version: 3.0.0 Summary (Device Total: 1) Progress: 60% 6 / 10 ET: 140s Rev: 9bfa978 |
181++---------------------------------------------------------------------------------------------------------+
182+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_7 OnGeirExecute RUNNING 1s |
183++---------------------------------------------------------------------------------------------------------+
184+2026-09-28 02:28:29,970 [INFO] [1794572 MP MT] [instance_base.py:301]:
185++----------------------------------------------------------------------------------------------------------+
186+| Version: 3.0.0 Summary (Device Total: 1) Progress: 60% 6 / 10 ET: 150s Rev: 9bfa978 |
187++----------------------------------------------------------------------------------------------------------+
188+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_7 OnGeirExecute RUNNING 11s |
189++----------------------------------------------------------------------------------------------------------+
190+2026-09-28 02:28:33,540 [INFO] [1794586 auto_testcase_name_7 OnReturning] [profiling.py:63]:
191+########################
192+Performance result: Comparison result:
193+CST_PERF: None CST_GOLD: PASS,PASS
194+DYN_PERF: None DYN_GOLD:
195+PRECISION_STATUS: PASS PASSED: PASS
196+########################
197+
198+2026-09-28 02:28:38,043 [INFO] [1794586 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
199+2026-09-28 02:28:38,043 [INFO] [1794586 auto_testcase_name_8 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_8.json
200+2026-09-28 02:28:38,043 [INFO] [1794586 auto_testcase_name_8 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
201+2026-09-28 02:28:39,977 [INFO] [1794572 MP MT] [instance_base.py:301]:
202++---------------------------------------------------------------------------------------------------------+
203+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 160s Rev: 9bfa978 |
204++---------------------------------------------------------------------------------------------------------+
205+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_8 OnGeirExecute RUNNING 1s |
206++---------------------------------------------------------------------------------------------------------+
207+2026-09-28 02:28:49,983 [INFO] [1794572 MP MT] [instance_base.py:301]:
208++----------------------------------------------------------------------------------------------------------+
209+| Version: 3.0.0 Summary (Device Total: 1) Progress: 70% 7 / 10 ET: 170s Rev: 9bfa978 |
210++----------------------------------------------------------------------------------------------------------+
211+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_8 OnGeirExecute RUNNING 11s |
212++----------------------------------------------------------------------------------------------------------+
213+2026-09-28 02:28:52,817 [INFO] [1794586 auto_testcase_name_8 OnReturning] [profiling.py:63]:
214+########################
215+Performance result: Comparison result:
216+CST_PERF: None CST_GOLD: PASS,PASS
217+DYN_PERF: None DYN_GOLD:
218+PRECISION_STATUS: PASS PASSED: PASS
219+########################
220+
221+2026-09-28 02:28:57,308 [INFO] [1794586 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
222+2026-09-28 02:28:57,309 [INFO] [1794586 auto_testcase_name_9 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_9.json
223+2026-09-28 02:28:57,309 [INFO] [1794586 auto_testcase_name_9 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
224+2026-09-28 02:28:59,993 [INFO] [1794572 MP MT] [instance_base.py:301]:
225++---------------------------------------------------------------------------------------------------------+
226+| Version: 3.0.0 Summary (Device Total: 1) Progress: 80% 8 / 10 ET: 180s Rev: 9bfa978 |
227++---------------------------------------------------------------------------------------------------------+
228+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_9 OnGeirExecute RUNNING 2s |
229++---------------------------------------------------------------------------------------------------------+
230+2026-09-28 02:29:09,997 [INFO] [1794572 MP MT] [instance_base.py:301]:
231++----------------------------------------------------------------------------------------------------------+
232+| Version: 3.0.0 Summary (Device Total: 1) Progress: 80% 8 / 10 ET: 190s Rev: 9bfa978 |
233++----------------------------------------------------------------------------------------------------------+
234+| NPU:0 Ascend950PR_9579 | 1794586 _testcase_name_9 OnGeirExecute RUNNING 12s |
235++----------------------------------------------------------------------------------------------------------+
236+2026-09-28 02:29:12,449 [INFO] [1794586 auto_testcase_name_9 OnReturning] [profiling.py:63]:
237+########################
238+Performance result: Comparison result:
239+CST_PERF: None CST_GOLD: PASS,PASS
240+DYN_PERF: None DYN_GOLD:
241+PRECISION_STATUS: PASS PASSED: PASS
242+########################
243+
244+2026-09-28 02:29:16,877 [INFO] [1794586 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:250]: GEIR op source unchanged: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta.cpp
245+2026-09-28 02:29:16,877 [INFO] [1794586 auto_testcase_name_10 OnGeirCompile] [graph_builder.py:625]: Generated GEIR case config: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/const/auto_testcase_name_10.json
246+2026-09-28 02:29:16,877 [INFO] [1794586 auto_testcase_name_10 OnGeirCompile] [compiler.py:188]: GEIR op binary cached: /workspace/A-September_changwei/INTrainingUpdateGradGammaBeta/ops-test-kit/geir/ops/in_training_update_grad_gamma_beta
247+2026-09-28 02:29:19,999 [INFO] [1794572 MP MT] [instance_base.py:301]:
248++---------------------------------------------------------------------------------------------------------+
249+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 200s Rev: 9bfa978 |
250++---------------------------------------------------------------------------------------------------------+
251+| NPU:0 Ascend950PR_9579 | 1794586 testcase_name_10 OnGeirExecute RUNNING 3s |
252++---------------------------------------------------------------------------------------------------------+
253+2026-09-28 02:29:30,006 [INFO] [1794572 MP MT] [instance_base.py:301]:
254++----------------------------------------------------------------------------------------------------------+
255+| Version: 3.0.0 Summary (Device Total: 1) Progress: 90% 9 / 10 ET: 210s Rev: 9bfa978 |
256++----------------------------------------------------------------------------------------------------------+
257+| NPU:0 Ascend950PR_9579 | 1794586 testcase_name_10 OnGeirExecute RUNNING 13s |
258++----------------------------------------------------------------------------------------------------------+
259+2026-09-28 02:29:31,659 [INFO] [1794586 auto_testcase_name_10 OnReturning] [profiling.py:63]:
260+########################
261+Performance result: Comparison result:
262+CST_PERF: None CST_GOLD: PASS,PASS
263+DYN_PERF: None DYN_GOLD:
264+PRECISION_STATUS: PASS PASSED: PASS
265+########################
266+
267+2026-09-28 02:29:31,668 [INFO] [1794572 MP MT] [instance_base.py:147]: ttk Profiling complete
268+2026-09-28 02:29:31,692 [INFO] [1794572 MP MT] [heartbeat_manager.py:158]: tenant c6a56ec8ab24 deleted on 127.0.0.1:19098
269+2026-09-28 02:29:31,704 [INFO] [1794572 MP MT] [instance_base.py:325]:
270+==================== TTK Test Summary ===================
271+ Total : 10
272+ PASS : 10
273+ FAIL : 0
274+ SKIP/NA : 0
275+ PassRate : 100.00% (10/10)
276+=========================================================
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/driver.log+113-0
@@ -0,0 +1,113 @@
1+2026-09-28 00:45:57 shared performance lock acquired
2+2026-09-28 00:45:58 resource not idle; queued
3+2026-09-28 00:46:30 chunk_00 attempt 1 START
4+2026-09-28 00:47:03 chunk_00 attempt 1 INTERRUPTED by resource gate
5+2026-09-28 00:47:03 chunk_00 attempt 1 discarded rc=143 invalid=1
6+2026-09-28 00:47:06 resource not idle; queued
7+2026-09-28 00:47:21 resource not idle; queued
8+2026-09-28 00:47:36 resource not idle; queued
9+2026-09-28 00:47:51 resource not idle; queued
10+2026-09-28 00:48:06 resource not idle; queued
11+2026-09-28 00:48:21 resource not idle; queued
12+2026-09-28 00:48:36 resource not idle; queued
13+2026-09-28 00:48:51 resource not idle; queued
14+2026-09-28 00:49:06 resource not idle; queued
15+2026-09-28 00:49:21 resource not idle; queued
16+2026-09-28 00:49:36 resource not idle; queued
17+2026-09-28 00:49:51 resource not idle; queued
18+2026-09-28 00:50:06 resource not idle; queued
19+2026-09-28 00:50:21 resource not idle; queued
20+2026-09-28 00:50:36 resource not idle; queued
21+2026-09-28 00:50:51 resource not idle; queued
22+2026-09-28 00:51:06 resource not idle; queued
23+2026-09-28 00:51:21 resource not idle; queued
24+2026-09-28 00:51:36 resource not idle; queued
25+2026-09-28 00:51:51 resource not idle; queued
26+2026-09-28 00:52:07 resource not idle; queued
27+2026-09-28 00:52:22 resource not idle; queued
28+2026-09-28 00:52:54 chunk_00 attempt 2 START
29+2026-09-28 00:54:30 chunk_00 attempt 2 INTERRUPTED by resource gate
30+2026-09-28 00:54:30 chunk_00 attempt 2 discarded rc=143 invalid=1
31+2026-09-28 00:54:33 resource not idle; queued
32+2026-09-28 00:54:49 resource not idle; queued
33+2026-09-28 00:55:05 resource not idle; queued
34+2026-09-28 00:55:22 resource not idle; queued
35+2026-09-28 00:55:38 resource not idle; queued
36+2026-09-28 01:02:34 shared performance lock acquired
37+2026-09-28 01:03:39 resource not idle; queued
38+2026-09-28 01:03:55 resource not idle; queued
39+2026-09-28 01:04:12 resource not idle; queued
40+2026-09-28 01:04:44 resource not idle; queued
41+2026-09-28 01:05:00 resource not idle; queued
42+2026-09-28 01:05:16 resource not idle; queued
43+2026-09-28 01:05:32 resource not idle; queued
44+2026-09-28 01:06:04 resource not idle; queued
45+2026-09-28 01:06:36 resource not idle; queued
46+2026-09-28 01:06:53 resource not idle; queued
47+2026-09-28 01:07:26 resource not idle; queued
48+2026-09-28 01:07:42 resource not idle; queued
49+2026-09-28 01:07:58 resource not idle; queued
50+2026-09-28 01:08:14 resource not idle; queued
51+2026-09-28 01:08:47 resource not idle; queued
52+2026-09-28 01:09:19 resource not idle; queued
53+2026-09-28 01:09:51 resource not idle; queued
54+2026-09-28 01:10:23 resource not idle; queued
55+2026-09-28 01:10:55 resource not idle; queued
56+2026-09-28 01:11:28 resource not idle; queued
57+2026-09-28 01:12:00 resource not idle; queued
58+2026-09-28 01:12:32 resource not idle; queued
59+2026-09-28 01:12:48 resource not idle; queued
60+2026-09-28 01:13:04 resource not idle; queued
61+2026-09-28 01:13:20 resource not idle; queued
62+2026-09-28 01:13:36 resource not idle; queued
63+2026-09-28 01:14:08 resource not idle; queued
64+2026-09-28 01:14:25 resource not idle; queued
65+2026-09-28 01:14:41 resource not idle; queued
66+2026-09-28 01:14:57 resource not idle; queued
67+2026-09-28 01:15:29 resource not idle; queued
68+2026-09-28 01:15:45 resource not idle; queued
69+2026-09-28 01:16:18 resource not idle; queued
70+2026-09-28 01:16:50 resource not idle; queued
71+2026-09-28 01:17:22 resource not idle; queued
72+2026-09-28 01:17:54 resource not idle; queued
73+2026-09-28 01:18:10 resource not idle; queued
74+2026-09-28 01:18:59 resource not idle; queued
75+2026-09-28 01:19:15 resource not idle; queued
76+2026-09-28 01:19:31 resource not idle; queued
77+2026-09-28 01:19:47 resource not idle; queued
78+2026-09-28 01:20:03 resource not idle; queued
79+2026-09-28 01:20:36 resource not idle; queued
80+2026-09-28 01:20:52 resource not idle; queued
81+2026-09-28 01:21:08 resource not idle; queued
82+2026-09-28 01:21:24 resource not idle; queued
83+2026-09-28 01:21:40 resource not idle; queued
84+2026-09-28 01:21:57 resource not idle; queued
85+2026-09-28 01:22:13 resource not idle; queued
86+2026-09-28 01:22:45 resource not idle; queued
87+2026-09-28 01:23:01 resource not idle; queued
88+2026-09-28 01:23:17 resource not idle; queued
89+2026-09-28 01:23:49 resource not idle; queued
90+2026-09-28 01:24:06 resource not idle; queued
91+2026-09-28 01:25:43 resource not idle; queued
92+2026-09-28 01:25:59 resource not idle; queued
93+2026-09-28 01:26:16 resource not idle; queued
94+2026-09-28 01:26:32 resource not idle; queued
95+2026-09-28 01:26:48 resource not idle; queued
96+2026-09-28 01:27:04 resource not idle; queued
97+2026-09-28 01:27:20 resource not idle; queued
98+2026-09-28 01:27:36 resource not idle; queued
99+2026-09-28 01:27:52 resource not idle; queued
100+2026-09-28 01:28:25 resource not idle; queued
101+2026-09-28 01:28:57 resource not idle; queued
102+2026-09-28 01:31:22 chunk_00 attempt 3 START
103+2026-09-28 01:34:50 chunk_00 attempt 3 audit failed; discarded
104+2026-09-28 02:06:48 shared performance lock acquired
105+2026-09-28 02:06:48 chunk_00 already accepted; retaining original evidence
106+2026-09-28 02:08:59 chunk_01 attempt 1 START
107+2026-09-28 02:12:27 chunk_01 attempt 1 ACCEPTED
108+2026-09-28 02:14:38 chunk_02 attempt 1 START
109+2026-09-28 02:18:08 chunk_02 attempt 1 ACCEPTED
110+2026-09-28 02:20:18 chunk_03 attempt 1 START
111+2026-09-28 02:23:48 chunk_03 attempt 1 ACCEPTED
112+2026-09-28 02:25:58 chunk_04 attempt 1 START
113+2026-09-28 02:29:44 chunk_04 attempt 1 ACCEPTED
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/extract_device_times.py+146-0
@@ -0,0 +1,146 @@
1+#!/usr/bin/env python3
2+"""Recover device times from this CANN's msprof SQLite when TTK CSV says UNKNOWN.
3+ 
4+Uses the same kernel task filter and warmup rule as TTK's rts_sequence.py.
5+Keep the raw TTK CSV and profiler DBs as primary evidence; this is an audit view.
6+"""
7+ 
8+import argparse
9+import ast
10+import csv
11+import datetime as dt
12+import json
13+import math
14+import pathlib
15+import sqlite3
16+import statistics
17+import shutil
18+import hashlib
19+import re
20+ 
21+ 
22+KERNEL_TYPES = {
23+ "AI_CORE", "AIV_SQE", "AI_VECTOR_CORE", "MIX_AIC", "MIX_AIV",
24+ "KERNEL_MIX_AIC", "KERNEL_MIX_AIV", "KERNEL_AIVEC", "KERNEL_AICORE",
25+}
26+ 
27+ 
28+def main():
29+ parser = argparse.ArgumentParser()
30+ parser.add_argument("--csv", required=True)
31+ parser.add_argument("--mode", choices=("kernel", "geir"), required=True)
32+ parser.add_argument("--since", required=True, help="local timestamp YYYY-MM-DD HH:MM:SS")
33+ parser.add_argument("--output", required=True)
34+ parser.add_argument("--summary", required=True)
35+ args = parser.parse_args()
36+ floor = dt.datetime.strptime(args.since, "%Y-%m-%d %H:%M:%S").timestamp()
37+ project = pathlib.Path(__file__).resolve().parents[2]
38+ profiler_root = project / "ops-test-kit" / "msprof" / ("op" if args.mode == "kernel" else "geir")
39+ 
40+ with open(args.csv, encoding="utf-8-sig", newline="") as stream:
41+ rows = list(csv.DictReader(stream))
42+ if not rows:
43+ raise ValueError("empty TTK CSV")
44+ names = [row["testcase_name"] for row in rows]
45+ if len(set(names)) != len(names):
46+ raise ValueError("duplicate testcase_name in TTK result")
47+ 
48+ results = []
49+ for row in rows:
50+ name = row["testcase_name"]
51+ if pathlib.Path(name).name != name or ".." in name:
52+ raise ValueError(f"unsafe testcase name: {name!r}")
53+ if row.get("precision_status") != "PASS":
54+ raise ValueError(f"precision did not pass: {name}")
55+ if args.mode == "kernel" and row.get("bin_compile_s") != "BINARY_MATCH":
56+ raise ValueError(f"binary was not matched: {name}")
57+ 
58+ metrics = ast.literal_eval(row.get("xpu_metrics") or "{}")
59+ gpu = metrics.get("torch", {})
60+ gpu_us = float(gpu.get("device_us", 0))
61+ if gpu.get("status") != "PASS" or not math.isfinite(gpu_us) or gpu_us <= 0:
62+ raise ValueError(f"valid GPU kernel time missing: {name}: {metrics}")
63+ 
64+ dbs = []
65+ for db in (profiler_root / name).glob("**/device_0/sqlite/ai_core_op_summary.db"):
66+ session = next((part for part in db.parts if part.startswith("PROF_")), "")
67+ match = re.fullmatch(r"PROF_\d+_(\d{17})_(\d{8})[A-Z]+", session)
68+ if not match:
69+ raise ValueError(f"unexpected profiler session identity: {session}")
70+ session_start = dt.datetime.strptime(match[1], "%Y%m%d%H%M%S%f").timestamp()
71+ # msprof export may regenerate an interrupted OLD session's DB.
72+ # A fresh DB mtime alone does not prove a fresh device launch.
73+ if session_start >= floor and db.stat().st_mtime >= floor and (args.mode == "geir" or "/binary/" in str(db)):
74+ if args.mode == "geir":
75+ pid = int(match[2])
76+ runtime_file = pathlib.Path(__file__).resolve().parent / f"runtime_libraries_{pid}.json"
77+ runtime = json.loads(runtime_file.read_text())
78+ if '/' + name + '_input ' not in runtime['cmdline']:
79+ raise ValueError(f"profiler PID does not map to this GEIR testcase: {pid}, {name}")
80+ dbs.append(db)
81+ if len(dbs) != 1:
82+ raise ValueError(f"expected one fresh profiler DB for {name}, found {len(dbs)}")
83+ db = dbs[0]
84+ original_db = db
85+ archive = pathlib.Path(args.output).resolve().parent / "profiler_sqlite" / name
86+ archive.mkdir(parents=True, exist_ok=True)
87+ archived_db = archive / db.name
88+ if archived_db.exists():
89+ assert hashlib.sha256(db.read_bytes()).digest() == hashlib.sha256(archived_db.read_bytes()).digest()
90+ else:
91+ shutil.copy2(db, archived_db)
92+ db = archived_db
93+ with sqlite3.connect(f"file:{db}?mode=ro", uri=True) as conn:
94+ tasks = conn.execute(
95+ "SELECT task_type, duration_time FROM task_time ORDER BY start_time, task_id"
96+ ).fetchall()
97+ durations_us = [float(ns) / 1000.0 for kind, ns in tasks if kind in KERNEL_TYPES]
98+ if args.mode == "kernel":
99+ # TTK rts_sequence._process_msprof_cycles ignores its first launch.
100+ if len(durations_us) < 2:
101+ raise ValueError(f"fewer than two kernel launches in {db}")
102+ measured = durations_us[1:]
103+ else:
104+ # TTK geir.profiling._parse_msprof_task_duration takes all launches.
105+ if not durations_us:
106+ raise ValueError(f"no kernel task duration in {db}")
107+ measured = durations_us
108+ npu_us = round(statistics.median(measured), 3)
109+ if npu_us <= 0:
110+ raise ValueError(f"invalid NPU kernel time: {name}")
111+ results.append({
112+ "testcase_name": name,
113+ "npu_us": npu_us,
114+ "gpu_us": gpu_us,
115+ "gpu_over_npu": round(gpu_us / npu_us, 6),
116+ "kernel_task_count": len(durations_us),
117+ "profiler_db": str(db),
118+ "original_profiler_db": str(original_db),
119+ "profiler_db_sha256": hashlib.sha256(db.read_bytes()).hexdigest(),
120+ "ttk_perf_field": row.get("bin_perf_us" if args.mode == "kernel" else "cst_perf_us", ""),
121+ })
122+ 
123+ output = pathlib.Path(args.output)
124+ with output.open("w", encoding="utf-8", newline="") as stream:
125+ writer = csv.DictWriter(stream, fieldnames=list(results[0]))
126+ writer.writeheader()
127+ writer.writerows(results)
128+ report = {
129+ "mode": args.mode,
130+ "case_count": len(results),
131+ "min_gpu_over_npu": min(item["gpu_over_npu"] for item in results),
132+ "ratio_le_0_1": [item["testcase_name"] for item in results if item["gpu_over_npu"] <= 0.1],
133+ "source_csv": str(pathlib.Path(args.csv).resolve()),
134+ "method": (
135+ "msprof ai_core_op_summary.db task_time duration_time ns/1000; "
136+ + ("kernel excludes first launch" if args.mode == "kernel" else "GEIR includes all kernel tasks")
137+ ),
138+ }
139+ pathlib.Path(args.summary).write_text(json.dumps(report, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
140+ print(json.dumps(report, ensure_ascii=False))
141+ if report["ratio_le_0_1"]:
142+ raise SystemExit("GPU/NPU ratio must be > 0.1 for every case")
143+ 
144+ 
145+if __name__ == "__main__":
146+ main()
Ahandover_evidence/INTrainingUpdateGradGammaBeta/geir_crosscheck_20260928/finalize.log+35-0
@@ -0,0 +1,35 @@
1+{"source_head": "41169c73072287e8def27bf3bc606fa16869b63b", "kernel_files_matched": 5}
2+{
3+ "case_count": 50,
4+ "cross_check_output_count": 100,
5+ "precision_pass": 50,
6+ "gpu_endpoint": "gpu-20102",
7+ "remote_service_port": 19199,
8+ "local_tunnel_port": 19098,
9+ "ttk_head": "9bfa978d8098e2c96fd33380abab0a3c9dfd120c",
10+ "gpu_environment_evidence": [
11+ "gpu_environment_20260927T171122Z.json",
12+ "gpu_environment_20260927T183024Z.json"
13+ ],
14+ "precision_fail": 0,
15+ "gpu_perf_valid": 50,
16+ "npu_sqlite_valid": 50,
17+ "min_gpu_over_npu": 0.842386,
18+ "ratio_le_0_1": [],
19+ "runtime_vendor_loading_cases": 50,
20+ "raw_ttk_cst_perf_us_counts": {
21+ "CST_OFF": 50
22+ },
23+ "check_invoke_path_exit_code": 0,
24+ "check_invoke_path_sha256": "a897c904fd71dde054f4ae142c1618f4bde53187128a5ed3790003d9f11252b6",
25+ "invalid_attempts": [
26+ "runs/chunk_00/attempt_01",
27+ "runs/chunk_00/attempt_02"
28+ ],
29+ "freshness_sha256": "064d52684d04d754474ff93176ac4fce7fe3540bd56b764f734a68b415a4ad92",
30+ "limitations": [
31+ "Private built-in registration patch is not a formal canndev source rebuild.",
32+ "TTK cst_perf_us is preserved verbatim; NPU device time is separately audited from fresh SQLite task_time ns/1000.",
33+ "Source HEAD labels the repository; no claim that vendor libraries were rebuilt from the Golden-only commit."
34+ ]
35+}