feat: add excel support-ops-test-kit-AtomGit

M.gitignore+35-23

++# IDE / Editor
 .idea/
--*.pyc
--*.log
--test.py
--**/*.log
--tests/**/*.html
--kernel_meta/
--models/
--/geir/
--model_results/
--**/*_result.csv
--*.bin
--*.zip
--*.o
--toml_data/
--model_*/
--models_*/
--lib/
--LLMAnalysisTool/
--settings.local.json
--ttk/_build_hash.py
++# Python
++*.pyc
++
++# Build artifacts
++*.o
++*.zip
 csrc/build/
 csrc/*/build/
--msprof/
++ttk/_build_hash.py
++
++# Logs
++**/*.log
 log/
--# xpu_server runtime: default sync/tmp dirs (created next to the server package)
++# TTK runtime outputs
++kernel_meta/
++/geir/
++msprof/
++sim_output/
++**/*_result.csv
++
++# Models & data
++*.bin
++models/
++model_*/
++models_*/
++toml_data/
++
++# Remote server runtime (default sync/tmp dirs next to the server package)
 ttk/remote/server/ttk_xpu_sync/
 ttk/remote/server/ttk_tmp_dir/
--sim_output/
++# Local config
++settings.local.json
++
++# Misc temp
++test.py
++tests/**/*.html
++lib/
++LLMAnalysisTool/
++examples/case_store/generate_xlsx_examples.py

MAGENTS.md+3-1

	@@ -18,7 +18,7 @@ TTK 的四种测试模式对应昇腾技术栈的不同层级，越往上覆盖
18	└──────────────────────────────────────────────────┘	18	└──────────────────────────────────────────────────┘
19	```	19	```
20		20
21	-不同层级使用不同的测试用例结构，模式选错会导致用例解析失败。根据 CSV 表头自动判断模式（详见各 skill）。	21	+不同层级使用不同的测试用例结构，模式选错会导致用例解析失败。根据表头自动判断模式（详见各 skill）。输入支持 CSV 与 Excel(.xlsx)，详见 [用例生成](./docs/Test_Case_Generation.md)。
22		22
23	## 使用方式	23	## 使用方式
24		24
	@@ -28,6 +28,8 @@ TTK 的四种测试模式对应昇腾技术栈的不同层级，越往上覆盖
28	- ACLNN 模式：`python3 -m ttk aclnn -i cases.csv`	28	- ACLNN 模式：`python3 -m ttk aclnn -i cases.csv`
29	- E2E 模式：`python3 -m ttk e2e -i cases.csv`（默认自动探测 NPU；`--cpu` 强制 CPU）	29	- E2E 模式：`python3 -m ttk e2e -i cases.csv`（默认自动探测 NPU；`--cpu` 强制 CPU）
30		30
		31	+> 输入支持 CSV 与 Excel(.xlsx)；Excel 默认读首个工作表，用 `--sheet` 指定，如 `ttk kernel -i cases.xlsx --sheet T2`。详见 [任务执行](./docs/Task_Execution.md)。
		32	+
31	验证 NPU 环境：`python3 -m ttk info`	33	验证 NPU 环境：`python3 -m ttk info`
32		34
33	## 技能索引	35	## 技能索引

Mdocs/Operator_Test_Guides/ACLNN_Case_Writing.md+17-2

	@@ -2,7 +2,7 @@
2		2
3	适用于 `python3 -m ttk aclnn`，使用 `TestcaseAclnn`，共 27 个字段（含[公共字段](../Test_Case_Generation.md#公共字段所有模式通用)）。	3	适用于 `python3 -m ttk aclnn`，使用 `TestcaseAclnn`，共 27 个字段（含[公共字段](../Test_Case_Generation.md#公共字段所有模式通用)）。
4		4
5	-## 身份标识	5	+## 用例标识
6		6
7	\| 字段 \| 类型 \| 是否必填 \| 默认值 \| 说明 \|	7	\| 字段 \| 类型 \| 是否必填 \| 默认值 \| 说明 \|
8	\|------\|------\|---------\|--------\|------\|	8	\|------\|------\|---------\|--------\|------\|
	@@ -23,7 +23,7 @@
23		23
24	\| 字段 \| 类型 \| 是否必填 \| 默认值 \| 说明 \|	24	\| 字段 \| 类型 \| 是否必填 \| 默认值 \| 说明 \|
25	\|------\|------\|---------\|--------\|------\|	25	\|------\|------\|---------\|--------\|------\|
26	-\| `output_tensor_indexes` \| INT_TUPLE \| 否 \| (自动) \| 指示哪些张量是输出的索引。未设置时按参数命名规则自动填充：以 `Ref`/`Out`/`Output`/`OutOptional`/`OutputOptional` 结尾或等于 `output` 的张量参数识别为输出；Backward/Grad API 会排除 `gradOutput`/`gradOut`/`grad_output`/`attentionOut`/`dOut` 以及非末位的 `output`；无匹配时回退到最后一个张量 \|	26	+\| `output_tensor_indexes` \| INT_TUPLE \| 否 \| (自动) \| 指示哪些张量是输出的索引。未设置时按参数命名规则自动填充：以 `Ref`/`Out`/`Output`/`OutOptional`/`OutputOptional` 结尾或等于 `output` 的张量参数识别为输出；Backward/Grad API 会排除 `gradOutput`/`gradOut`/`grad_output`/`attentionOut` 以及非末位的 `output`；无匹配时回退到最后一个张量 \|
27	\| `output_inplace_indexes` \| INT_TUPLE \| 否 \| `()` \| 原地操作输出索引。无需手动填写：框架自动从算子信息库推断（当输出名称与输入名称相同时，表示输出覆写到该输入内存上） \|	27	\| `output_inplace_indexes` \| INT_TUPLE \| 否 \| `()` \| 原地操作输出索引。无需手动填写：框架自动从算子信息库推断（当输出名称与输入名称相同时，表示输出覆写到该输入内存上） \|
28		28
29	## 属性与标量	29	## 属性与标量
	@@ -94,3 +94,18 @@ full,aclnnAdd,"((10,8),)",(100,),,,...
94	```	94	```
95		95
96	上例中 `slice_0` 取输出轴 0 的 `[0:5]`，`slice_1` 取 `[5:10]`，`full` 不切片取全部。三者 seed 相同，`slice_0` + `slice_1` 的切片长度之和等于 `full`，归入同一比对组。	96	上例中 `slice_0` 取输出轴 0 的 `[0:5]`，`slice_1` 取 `[5:10]`，`full` 不切片取全部。三者 seed 相同，`slice_0` + `slice_1` 的切片长度之和等于 `full`，归入同一比对组。
		97	+
		98	+## 参考用例
		99	+
		100	+`examples/case_store/aclnn/` 目录下的示例：
		101	+
		102	+\| 文件 \| 验证特性 \| 关键列 \|
		103	+\|------\|---------\|--------\|
		104	+\| `aclnn_add.csv` \| 基础 aclnn API + 标量参数 \| `scalar_dtypes` \|
		105	+\| `aclnn_cat.csv` \| 拼接 `dim` 属性 × 6 dtype \| `attributes`（dim）、`output_tensor_indexes` \|
		106	+\| `aclnn_convolution.csv` \| 复杂属性（stride/padding/dilation/groups）解析 \| `attributes`、`tensor_formats` \|
		107	+\| `aclnn_inplace_fill_tensor.csv` \| 原地操作（inplace） \| `input_data_ranges`、`precision_tolerances` \|
		108	+\| `aclnn_masked_select.csv` \| bool mask 筛选 + 动态输出（elewise/broadcast） \| `input_data_ranges`、`precision_tolerances` \|
		109	+\| `aclnn_nonzero_v2.csv` \| `as_tuple` 属性分支 + 动态输出 \| `attributes`（as_tuple） \|
		110	+\| `aclnn_split_tensor.csv` \| 拆分为 TensorList 输出 + `splitSections`/`dim` \| `attributes`、`output_tensor_indexes` \|
		111	+\| `aclnn_add.xlsx` \| xlsx 多 sheet（T1/T2）输入验证 \| — \|

Mdocs/Operator_Test_Guides/ACLNN_Test_Guide.md+4-0

	@@ -157,6 +157,10 @@ python3 -m ttk aclnn -i examples/case_store/aclnn/aclnn_inplace_fill_tensor.csv
157		157
158	# 使用自定义Golden插件	158	# 使用自定义Golden插件
159	python3 -m ttk aclnn -i aclnn_cat.csv --plugin /path/to/my_golden.py	159	python3 -m ttk aclnn -i aclnn_cat.csv --plugin /path/to/my_golden.py
		160	+
		161	+# Excel 多 sheet 用例（默认首个工作表；--sheet 指定工作表）
		162	+python3 -m ttk aclnn -i examples/case_store/aclnn/aclnn_add.xlsx
		163	+python3 -m ttk aclnn -i examples/case_store/aclnn/aclnn_add.xlsx --sheet T2
160	```	164	```
161		165
162	> 通用参数（用例筛选/设备并行/精度控制/调试/结果输出）见[任务执行](../Task_Execution.md)	166	> 通用参数（用例筛选/设备并行/精度控制/调试/结果输出）见[任务执行](../Task_Execution.md)

Mdocs/Operator_Test_Guides/E2E_Case_Writing.md+12-1

	@@ -12,7 +12,7 @@
12	\| `torch.nn.functional.relu` \| 子模块函数 \| 调用子模块中的函数 \|	12	\| `torch.nn.functional.relu` \| 子模块函数 \| 调用子模块中的函数 \|
13	\| `torch.Tensor.relu_` \| Tensor方法 \| 通过Tensor实例调用（原地操作） \|	13	\| `torch.Tensor.relu_` \| Tensor方法 \| 通过Tensor实例调用（原地操作） \|
14		14
15	-## 身份标识	15	+## 用例标识
16		16
17	\| 字段 \| 类型 \| 是否必填 \| 默认值 \| 说明 \|	17	\| 字段 \| 类型 \| 是否必填 \| 默认值 \| 说明 \|
18	\|------\|------\|---------\|--------\|------\|	18	\|------\|------\|---------\|--------\|------\|
	@@ -79,3 +79,14 @@ full,torch.add,"((10,8),)",(100,),,,...
79	```	79	```
80		80
81	上例中 `slice_0` 取输出轴 0 的 `[0:5]`，`slice_1` 取 `[5:10]`，`full` 不切片取全部。三者 seed 相同，`slice_0` + `slice_1` 的切片长度之和等于 `full`，归入同一比对组。	81	上例中 `slice_0` 取输出轴 0 的 `[0:5]`，`slice_1` 取 `[5:10]`，`full` 不切片取全部。三者 seed 相同，`slice_0` + `slice_1` 的切片长度之和等于 `full`，归入同一比对组。
		82	+
		83	+## 参考用例
		84	+
		85	+`examples/case_store/e2e/` 目录下的示例：
		86	+
		87	+\| 文件 \| 框架 \| 验证特性 \| 关键列 \|
		88	+\|------\|------\|---------\|--------\|
		89	+\| `tf_ops.csv` \| TensorFlow \| 多 API（tf.raw_ops/nn/math/linalg）+ 多算子 \| `tensor_view_shapes`、`attributes` \|
		90	+\| `torch_add.csv` \| torch \| add/abs/relu/mm + inplace（`relu_`）/out 变体 + alpha 属性 \| `attributes`、`output_tensor_indexes` \|
		91	+\| `torch_npu_conv2d.csv` \| torch_npu \| NPU 专属 API + 自定义 golden \| `golden_api`、`output_tensor_indexes` \|
		92	+\| `torch_add.xlsx` \| torch \| xlsx 多 sheet（T1/T2）输入验证 \| — \|

Mdocs/Operator_Test_Guides/E2E_Test_Guide.md+4-0

	@@ -177,6 +177,10 @@ python3 -m ttk e2e -i examples/case_store/e2e/torch_add.csv --cpu
177		177
178	# 输出结果	178	# 输出结果
179	python3 -m ttk e2e -i torch_add.csv -o results.csv	179	python3 -m ttk e2e -i torch_add.csv -o results.csv
		180	+
		181	+# Excel 多 sheet 用例（默认首个工作表；--sheet 指定工作表）
		182	+python3 -m ttk e2e -i examples/case_store/e2e/torch_add.xlsx
		183	+python3 -m ttk e2e -i examples/case_store/e2e/torch_add.xlsx --sheet T2
180	```	184	```
181		185
182	> 通用参数（用例筛选/设备并行/精度控制/调试/结果输出）见[任务执行](../Task_Execution.md)	186	> 通用参数（用例筛选/设备并行/精度控制/调试/结果输出）见[任务执行](../Task_Execution.md)

Mdocs/Operator_Test_Guides/GEIR_Case_Writing.md+6-5

	@@ -21,9 +21,10 @@ Kernel 模式的 26 个字段说明详见 [Kernel用例编写](./Kernel_Case_Wri
21		21
22	## 参考用例	22	## 参考用例
23		23
24	-GEIR 复用 Kernel 的 CSV 文件，可直接用 `examples/case_store/kernel/` 下的用例。如需精确控制动态维度，补充 `dyn_input_shapes` 列即可：	24	+GEIR 复用 Kernel 的 CSV 用例（字段与验证场景见 [Kernel 参考用例](./Kernel_Case_Writing.md)），额外用 `dyn_input_shapes` 列控制图编译时的动态维度。本通路独立示例：
25		25
26	-\| 场景 \| 说明 \|	26	+\| 文件 \| 验证特性 \| 关键列 \|
27	-\|------\|------\|	27	+\|------\|---------\|--------\|
28	-\| 基本用例 \| 直接使用 `examples/case_store/kernel/` 下的 CSV，无需改动 \|	28	+\| `geir/add.xlsx` \| xlsx 多 sheet（T1/T2）输入 + `dyn_input_shapes` 图编译动态 shape \| `dyn_input_shapes` \|
29	-\| 精确控制动态维度 \| 在 CSV 末尾增加 `dyn_input_shapes` 列，指定各输入维度的动态性 \|	29	+
		30	+> 直接使用 `examples/case_store/kernel/` 下的 CSV 跑 GEIR 无需改动；如需精确控制某输入维度的动态性，在末尾追加 `dyn_input_shapes` 列即可。

Mdocs/Operator_Test_Guides/GEIR_Test_Guide.md+4-0

	@@ -186,6 +186,10 @@ python3 -m ttk geir -i add.csv --plugin /path/to/my_golden.py
186		186
187	# 重跑精度失败的用例	187	# 重跑精度失败的用例
188	python3 -m ttk geir -i add.csv --rerun=precision_status	188	python3 -m ttk geir -i add.csv --rerun=precision_status
		189	+
		190	+# Excel 多 sheet 用例（默认首个工作表；--sheet 指定工作表）
		191	+python3 -m ttk geir -i examples/case_store/geir/add.xlsx
		192	+python3 -m ttk geir -i examples/case_store/geir/add.xlsx --sheet T2
189	```	193	```
190		194
191	> 通用参数（用例筛选/设备并行/精度控制/调试/结果输出）见[任务执行](../Task_Execution.md)	195	> 通用参数（用例筛选/设备并行/精度控制/调试/结果输出）见[任务执行](../Task_Execution.md)

Mdocs/Operator_Test_Guides/Kernel_Case_Writing.md+11-11

	@@ -11,7 +11,7 @@
11	- 自定义算子信息库路径：`{ASCEND_OPP_PATH}/vendors/customize/op_impl/ai_core/tbe/config/{芯片系列号}/aic-{芯片系列号}-ops-info.json`	11	- 自定义算子信息库路径：`{ASCEND_OPP_PATH}/vendors/customize/op_impl/ai_core/tbe/config/{芯片系列号}/aic-{芯片系列号}-ops-info.json`
12	- 例如 add 算子在 `ascend910b` 芯片上的 builtin 信息库：`$ASCEND_OPP_PATH/built-in/op_impl/ai_core/tbe/config/ascend910b/ops_math/aic-ascend910b-ops-info-math.json`	12	- 例如 add 算子在 `ascend910b` 芯片上的 builtin 信息库：`$ASCEND_OPP_PATH/built-in/op_impl/ai_core/tbe/config/ascend910b/ops_math/aic-ascend910b-ops-info-math.json`
13		13
14	-## 身份标识	14	+## 用例标识
15		15
16	\| 字段 \| 类型 \| 是否必填 \| 默认值 \| 说明 \|	16	\| 字段 \| 类型 \| 是否必填 \| 默认值 \| 说明 \|
17	\|------\|------\|---------\|--------\|------\|	17	\|------\|------\|---------\|--------\|------\|
	@@ -62,13 +62,13 @@
62		62
63	`examples/case_store/kernel/` 目录下提供了各种场景的示例：	63	`examples/case_store/kernel/` 目录下提供了各种场景的示例：
64		64
65	-\| 文件 \| 涵盖场景 \|	65	+\| 文件 \| 验证特性 \| 关键列 \|
66	-\|------\|----------\|	66	+\|------\|---------\|--------\|
67	-\| `abs.csv` \| 基本用例，多 dtype \|	67	+\| `abs.csv` \| 一元算子 dtype 全覆盖（int8/16/32/64、fp16/32）+ 精度 \| 标准输入/输出列 \|
68	-\| `add.csv` \| 多 dtype、广播、input_data_ranges \|	68	+\| `add.csv` \| 二元 broadcast：6 dtype × 多 rank/广播模式 \| `input_data_ranges` 控制随机数据范围 \|
69	-\| `concat_d.csv` \| TensorList 输入（DYNAMIC） \|	69	+\| `concat_d.csv` \| 动态输入个数 + 负 `concat_dim` 拼接 \| `attributes`（concat_dim） \|
70	-\| `mat_mul_v3.csv` \| 可选输入（None 占位）、属性 \|	70	+\| `mat_mul_v3.csv` \| 矩乘 `transpose_x1/x2` 组合 × 15 组 shape + 可选输入（None 占位） \| `attributes`、`network_name` \|
71	-\| `non_zero.csv` \| 输出 shape 未知（output_shape_unknown_indexes） \|	71	+\| `non_zero.csv` \| 输出 shape 由输入取值决定（数据相关动态输出） \| `output_shape_unknown_indexes` 声明动态输出 \|
72	-\| `reduce_min.csv` \| ValueDepend 张量输入（axes 通过 attributes 指定） \|	72	+\| `reduce_min.csv` \| 规约 `axes`/`keepdims` + ValueDepend 张量输入 \| `attributes`（axes/keepdims） \|
73	-\| `split.csv` \| TensorList 输出、动态参数 \|	73	+\| `split.csv` \| 拆分为 TensorList 输出 + `split_dim`/`num_split` \| `attributes`，输出为多张量 \|
74	-\| `zeros_like.csv` \| 基本用例 \|	74	+\| `zeros_like.csv` \| 输出与输入同 shape 的零填充（无输入数据依赖） \| `output_shapes` 跟随 `input_shapes` \|

Mdocs/Operator_Test_Guides/Kernel_Test_Guide.md+4-0

	@@ -234,6 +234,10 @@ python3 -m ttk kernel -i add.csv --rerun=precision_status
234		234
235	# 使用自定义Golden插件	235	# 使用自定义Golden插件
236	python3 -m ttk kernel -i add.csv --plugin /path/to/my_golden.py	236	python3 -m ttk kernel -i add.csv --plugin /path/to/my_golden.py
		237	+
		238	+# Excel 多 sheet 用例（默认首个工作表；--sheet 指定工作表）
		239	+python3 -m ttk kernel -i examples/case_store/kernel/add.xlsx
		240	+python3 -m ttk kernel -i examples/case_store/kernel/add.xlsx --sheet T2
237	```	241	```
238		242
239	> 通用参数（用例筛选/设备并行/精度控制/调试/结果输出）见[任务执行](../Task_Execution.md)	243	> 通用参数（用例筛选/设备并行/精度控制/调试/结果输出）见[任务执行](../Task_Execution.md)

Mdocs/Task_Execution.md+6-2

	@@ -13,7 +13,7 @@ python3 -m ttk {kernel,aclnn,e2e,geir,info,list} [选项]
13	\| `aclnn` \| aclnn\* C API 调用 + 精度比对 \|	13	\| `aclnn` \| aclnn\* C API 调用 + 精度比对 \|
14	\| `e2e` \| 框架 API 端到端测试 \|	14	\| `e2e` \| 框架 API 端到端测试 \|
15	\| `info` \| 查询本机 Ascend NPU 设备信息 \|	15	\| `info` \| 查询本机 Ascend NPU 设备信息 \|
16	-\| `list` \| 预览 CSV 中的测试用例列表 \|	16	+\| `list` \| 预览用例文件中的测试用例列表 \|
17		17
18	查看版本：	18	查看版本：
19		19
	@@ -41,13 +41,17 @@ python3 -m ttk list -i examples/case_store/kernel/add.csv
41		41
42	# 按算子名过滤预览	42	# 按算子名过滤预览
43	python3 -m ttk list -i cases.csv --op add	43	python3 -m ttk list -i cases.csv --op add
		44	+
		45	+# 预览 Excel 用例（指定工作表）
		46	+python3 -m ttk list -i cases.xlsx --sheet Sheet2
44	```	47	```
45		48
46	# 通用参数	49	# 通用参数
47		50
48	\| 参数 \| 缩写 \| 默认值 \| 说明 \| 示例 \|	51	\| 参数 \| 缩写 \| 默认值 \| 说明 \| 示例 \|
49	\|------\|------\|--------\|------\|------\|	52	\|------\|------\|--------\|------\|------\|
50	-\| `--input` \| `-i` \| 无 \| CSV 用例文件路径（必填） \| `-i add.csv` \|	53	+\| `--input` \| `-i` \| 无 \| 用例文件路径，支持 csv/xlsx（必填） \| `-i add.csv` 或 `-i cases.xlsx` \|
		54	+\| `--sheet` \| \| 首个工作表 \| Excel 工作表名；仅 `.xlsx` 生效，csv 忽略。xlsx 默认输出名带实际 sheet 名（如 `cases_T2_result.csv`），多 sheet 互不覆盖 \| `--sheet T2` \|
51	\| `--config` \| \| 无 \| ttk 配置 YAML 路径（覆盖 `~/.config/ttk/` 和 `./ttk.conf.yaml`） \| `--config ttk.conf.yaml` \|	55	\| `--config` \| \| 无 \| ttk 配置 YAML 路径（覆盖 `~/.config/ttk/` 和 `./ttk.conf.yaml`） \| `--config ttk.conf.yaml` \|
52		56
53	## 用例筛选	57	## 用例筛选

Mdocs/Test_Case_Generation.md+26-1

	@@ -5,7 +5,7 @@
5		5
6	# 简介	6	# 简介
7		7
8	-TTK使用CSV文件批量定义测试用例。第一行为表头（列名），后续每行代表一个测试用例。不同测试模式使用不同的列定义。	8	+TTK使用表格文件（CSV 或 Excel .xlsx）批量定义测试用例。第一行为表头（列名），后续每行代表一个测试用例。不同测试模式使用不同的列定义。
9		9
10	模式自动识别逻辑（基于CSV表头 + 子命令）：	10	模式自动识别逻辑（基于CSV表头 + 子命令）：
11	- 表头包含 `api_name`：第一个非空 `api_name` 值不以 `aclnn` 开头 → E2E模式；以 `aclnn` 开头或全部为空 → ACLNN模式	11	- 表头包含 `api_name`：第一个非空 `api_name` 值不以 `aclnn` 开头 → E2E模式；以 `aclnn` 开头或全部为空 → ACLNN模式
	@@ -15,6 +15,31 @@ TTK使用CSV文件批量定义测试用例。第一行为表头（列名），
15		15
16	---	16	---
17		17
		18	+# 输入文件格式（CSV / Excel）
		19	+
		20	+TTK 支持两种等价的输入文件格式，表头与字段定义完全相同：
		21	+
		22	+\| 格式 \| 后缀 \| 多工作表 \| 说明 \|
		23	+\|------\|------\|---------\|------\|
		24	+\| CSV \| `.csv` \| — \| 文本表格，逗号分隔；含特殊字符（括号、逗号、引号）的字段须用双引号包裹。 \|
		25	+\| Excel \| `.xlsx` \| 支持 \| 工作簿，每个工作表（Sheet）是一张表。 \|
		26	+
		27	+## Excel（.xlsx）用法
		28	+
		29	+- 默认读取第一个工作表；用 `--sheet` 指定工作表名，不存在时报错并列出可用工作表：
		30	+ ```shell
		31	+ python3 -m ttk kernel -i cases.xlsx # 首个工作表
		32	+ python3 -m ttk kernel -i cases.xlsx --sheet T2 # 指定工作表
		33	+ ```
		34	+- 单元格一律按文本处理：读取时每个单元格被转为字符串并去除首尾空白，与 CSV 单元格行为完全一致，后续字段解析逻辑（`input_shapes`、`attributes` 等的 `eval`）原样复用。因此 xlsx 与 csv 的用例可互换，模式自动识别、字段回退、TensorList 嵌套等规则完全相同。
		35	+- 数值列建议设为文本格式：Excel 会自动推断数字类型（如把 `01` 存为 `1`、`1e-8` 存为浮点）。为保证与 CSV 完全一致，建议把 `input_shapes`/`input_dtypes`/`attributes`/`input_data_ranges`/`output_shapes` 等列的单元格格式预设为「文本」。
		36	+- 空行处理：整行空白的行会被跳过（与 CSV 空行一致）。
		37	+- 结果文件仍为 CSV：无论输入是 csv 还是 xlsx，结果输出始终是 CSV，便于 diff 与版本管理。默认命名：csv 为 `{stem}_result.csv`；xlsx 为 `{stem}_{sheet}_result.csv`（`sheet` 为实际读取的工作表名，默认首页时也带上首页名，如 `cases_T1_result.csv`），同一文件多 sheet 跑测互不覆盖。
		38	+
		39	+> CSV 中「含特殊字符的字段须双引号包裹」的规则不适用于 xlsx——Excel 单元格天然支持逗号、引号等字符，无需转义。
		40	+
		41	+---
		42	+
18	# 公共字段（所有模式通用）	43	# 公共字段（所有模式通用）
19		44
20	以下9个字段为Kernel、GEIR、ACLNN、E2E四种模式共有。	45	以下9个字段为Kernel、GEIR、ACLNN、E2E四种模式共有。

Mexamples/case_store/README.md+19-1

	@@ -1,4 +1,22 @@
1	### 目的	1	### 目的
2		2
3	- 1个算子1个用例，用于用例书写借鉴。	3	- 1个算子1个用例，用于用例书写借鉴。
4	-- 作为门槛，TTK任何改动必须执行这里的用例。	4	+- 作为门槛，TTK任何改动必须执行这里的用例。
		5	+- xlsx 示例验证 CSV 与 Excel 输入等价、多 sheet 选择（`--sheet`）与默认输出命名。
		6	+
		7	+> 各通路每个用例的「验证特性 + 关键列」详见对应 Case Writing 指南：
		8	+> [Kernel](../../docs/Operator_Test_Guides/Kernel_Case_Writing.md) ·
		9	+> [GEIR](../../docs/Operator_Test_Guides/GEIR_Case_Writing.md) ·
		10	+> [ACLNN](../../docs/Operator_Test_Guides/ACLNN_Case_Writing.md) ·
		11	+> [E2E](../../docs/Operator_Test_Guides/E2E_Case_Writing.md)。
		12	+
		13	+### Excel（.xlsx）多 sheet 示例
		14	+
		15	+每通路一个 2-sheet 工作簿（T1/T2），验证 xlsx 输入与 `--sheet` 切换（默认首个工作表；默认输出名带实际 sheet 名，如 `add_T2_result.csv`，多 sheet 互不覆盖）：
		16	+
		17	+运行示例：
		18	+
		19	+```shell
		20	+python3 -m ttk kernel -i examples/case_store/kernel/add.xlsx # 默认首个 sheet
		21	+python3 -m ttk kernel -i examples/case_store/kernel/add.xlsx --sheet T2
		22	+```

Aexamples/case_store/aclnn/aclnn_add.xlsx+0-0

Binary files do not support preview

Aexamples/case_store/e2e/torch_add.xlsx+0-0

Binary files do not support preview

Aexamples/case_store/geir/add.xlsx+0-0

Binary files do not support preview

Aexamples/case_store/kernel/add.xlsx+0-0

Binary files do not support preview

Mpyproject.toml+4-0

	@@ -32,6 +32,7 @@ dependencies = [
32	"six>=1.15",	32	"six>=1.15",
33	"sympy",	33	"sympy",
34	"pandas",	34	"pandas",
		35	+ "openpyxl>=3.1",
35	]	36	]
36		37
37	[project.optional-dependencies]	38	[project.optional-dependencies]
	@@ -86,6 +87,9 @@ extend-exclude = [
86	select = ["E", "F", "W", "I", "UP", "B", "FA"]	87	select = ["E", "F", "W", "I", "UP", "B", "FA"]
87	ignore = ["E501", "FA100"]	88	ignore = ["E501", "FA100"]
88		89
		90	+[tool.ruff.lint.per-file-ignores]
		91	+"ttk/utilities/__init__.py" = ["F401", "F403"]
		92	+
89	[tool.mypy]	93	[tool.mypy]
90	python_version = "3.8"	94	python_version = "3.8"
91	ignore_missing_imports = true	95	ignore_missing_imports = true

Mrequirements.txt+1-0

	@@ -6,3 +6,4 @@ decorator
6	six	6	six
7	clang	7	clang
8	jinja2	8	jinja2
		9	+openpyxl>=3.1

Mtests/cli/test_four_modes_entry.py+7-1

Mttk/cli/bridge.py+44-42

	@@ -9,6 +9,8 @@ _FLOAT_CLEAN_VALUE = re.compile(r"[+-]?(?:(?:[0-9]+\.[0-9]*\|\.[0-9]+\|[0-9]+)(?:[
9		9
10	def _apply_io_args(sw, args):	10	def _apply_io_args(sw, args):
11	sw.input_files = [args.input]	11	sw.input_files = [args.input]
		12	+ if getattr(args, "sheet", None):
		13	+ sw.sheet = args.sheet
12	if hasattr(args, "append_file") and args.append_file:	14	if hasattr(args, "append_file") and args.append_file:
13	sw.output_file_name = args.append_file	15	sw.output_file_name = args.append_file
14	sw.append_mode = True	16	sw.append_mode = True
	@@ -321,57 +323,57 @@ def _log_manual_data_configuration(sw):
321	)	323	)
322		324
323		325
324	-def _detect_framework_from_csv(input_files):	326	+def _detect_framework_from_csv(input_files, sheet=None):
325	- """Peek at the first CSV to detect framework from api_name column.	327	+ """Peek at the first table to detect framework from api_name column.
326		328
327	- Reads the CSV header to find the api_name column, then checks the first	329	+ Reads the table header (CSV or XLSX via the unified table reader) to
328	- data row's api_name value. Returns 'tf' if it starts with 'tf.' or	330	+ find the api_name column, then checks the first data row's api_name
329	- 'tensorflow.', otherwise 'torch'.	331	+ value. Returns 'tf' if it starts with 'tf.' or 'tensorflow.',
		332	+ otherwise 'torch'.
330		333
331	- One CSV must contain only one framework's APIs: torch_npu and npu_device	334	+ One table must contain only one framework's APIs: torch_npu and
332	- each initialize the NPU runtime exclusively, so mixing frameworks in a	335	+ npu_device each initialize the NPU runtime exclusively, so mixing
333	- single run causes runtime conflicts. This function detects the framework	336	+ frameworks in a single run causes runtime conflicts. This function
334	- from the first data row and validates that all subsequent rows are	337	+ detects the framework from the first data row and validates that all
335	- consistent.	338	+ subsequent rows are consistent.
336	"""	339	"""
337	if not input_files:	340	if not input_files:
338	return "torch"	341	return "torch"
339	- import csv
340		342
341	from ttk.core_modules.framework_api.framework_detector import detect_framework	343	from ttk.core_modules.framework_api.framework_detector import detect_framework
		344	+ from ttk.utilities.table_reader import read_table
342		345
343	try:	346	try:
344	- with open(input_files[0], newline="") as f:	347	+ header, rows = read_table(input_files[0], sheet)
345	- reader = csv.reader(f)
346	- header = next(reader, None)
347	- if not header:
348	- return "torch"
349	- try:
350	- api_idx = header.index("api_name")
351	- except ValueError:
352	- return "torch"
353	- row = next(reader, None)
354	- if not row or api_idx >= len(row):
355	- return "torch"
356	- first_api = row[api_idx].strip()
357	- first_framework = detect_framework(first_api)
358	- for row in reader:
359	- if api_idx < len(row):
360	- row_api = row[api_idx].strip()
361	- row_framework = detect_framework(row_api)
362	- if row_framework != first_framework:
363	- raise ValueError(
364	- f"Mixed frameworks in one CSV is not supported: "
365	- f"first row is {first_framework} (api_name='{first_api}'), "
366	- f"but found {row_framework} (api_name='{row_api}') in a later row. "
367	- f"Please split into separate CSV files per framework."
368	- )
369	- return first_framework
370	- except ValueError:
371	- raise
372	except Exception as e:	348	except Exception as e:
373	- logging.warning(f"Failed to detect framework from CSV, defaulting to torch: {e}")	349	+ logging.warning(f"Failed to detect framework from input, defaulting to torch: {e}")
374	- return "torch"	350	+ return "torch"
		351	+
		352	+ try:
		353	+ api_idx = header.index("api_name")
		354	+ except ValueError:
		355	+ return "torch"
		356	+
		357	+ first_api = None
		358	+ for row in rows:
		359	+ if api_idx < len(row) and row[api_idx]:
		360	+ first_api = row[api_idx]
		361	+ break
		362	+ if not first_api:
		363	+ return "torch"
		364	+
		365	+ first_framework = detect_framework(first_api)
		366	+ for row in rows:
		367	+ if api_idx < len(row):
		368	+ row_framework = detect_framework(row[api_idx])
		369	+ if row_framework != first_framework:
		370	+ raise ValueError(
		371	+ f"Mixed frameworks in one table is not supported: "
		372	+ f"first row is {first_framework} (api_name='{first_api}'), "
		373	+ f"but found {row_framework} (api_name='{row[api_idx]}') in a later row. "
		374	+ f"Please split into separate files per framework."
		375	+ )
		376	+ return first_framework
375		377
376		378
377	def _preload_plugin_modules(sw):	379	def _preload_plugin_modules(sw):
	@@ -422,7 +424,7 @@ def run_with_switches(sw):
422	if sw.test_mode == "framework-api":	424	if sw.test_mode == "framework-api":
423	from ttk.core_modules.framework_api.instance import FrameworkApiInstance	425	from ttk.core_modules.framework_api.instance import FrameworkApiInstance
424		426
425	- sw.framework = _detect_framework_from_csv(sw.input_files)	427	+ sw.framework = _detect_framework_from_csv(sw.input_files, getattr(sw, "sheet", None))
426	ins = FrameworkApiInstance()	428	ins = FrameworkApiInstance()
427	elif sw.test_mode == "geir":	429	elif sw.test_mode == "geir":
428	from ttk.core_modules.geir.instance import GeirInstance	430	from ttk.core_modules.geir.instance import GeirInstance

Mttk/cli/common.py+10-5

	@@ -1,17 +1,22 @@
1	-import argparse
2	-
3	from ttk.remote import is_remote_configured	1	from ttk.remote import is_remote_configured
4		2
5		3
6	def _add_io_args(parser):	4	def _add_io_args(parser):
7	- parser.add_argument("-i", "--input", required=True, help="CSV test case file")	5	+ parser.add_argument("-i", "--input", required=True, help="Test case file (csv/xlsx)")
		6	+ parser.add_argument(
		7	+ "--sheet", default=None, help="Excel worksheet name (default: first worksheet); ignored for csv"
		8	+ )
8	parser.add_argument(	9	parser.add_argument(
9	"--config", default=None, help="Path to ttk config YAML (overrides ~/.config/ttk/ and ./ttk.conf.yaml)"	10	"--config", default=None, help="Path to ttk config YAML (overrides ~/.config/ttk/ and ./ttk.conf.yaml)"
10	)	11	)
11	output_group = parser.add_mutually_exclusive_group()	12	output_group = parser.add_mutually_exclusive_group()
12	output_group.add_argument("-o", "--output", help="Output CSV file (overwrite existing)")	13	output_group.add_argument("-o", "--output", help="Output CSV file (overwrite existing)")
13	- output_group.add_argument("-a", "--append", dest="append_file", help="Append results to existing CSV file; "	14	+ output_group.add_argument(
14	- "overwrites if file header does not match")	15	+ "-a",
		16	+ "--append",
		17	+ dest="append_file",
		18	+ help="Append results to existing CSV file; overwrites if file header does not match",
		19	+ )
15		20
16		21
17	def _add_case_filter_args(parser):	22	def _add_case_filter_args(parser):

Mttk/cli/list_cmd.py+6-5

	@@ -1,18 +1,19 @@
1	def register_list_command(subparsers):	1	def register_list_command(subparsers):
2	- parser = subparsers.add_parser("list", help="List test cases from CSV file")	2	+ parser = subparsers.add_parser("list", help="List test cases from a test case file (csv/xlsx)")
3	- parser.add_argument("-i", "--input", required=True, help="CSV test case file")	3	+ parser.add_argument("-i", "--input", required=True, help="Test case file (csv/xlsx)")
		4	+ parser.add_argument("--sheet", default=None, help="Excel worksheet name (default: first worksheet)")
4	parser.add_argument("--op", "--operator", dest="operator", help="Filter by operator name")	5	parser.add_argument("--op", "--operator", dest="operator", help="Filter by operator name")
5	parser.set_defaults(handler=_handle_list)	6	parser.set_defaults(handler=_handle_list)
6		7
7		8
8	def _handle_list(args):	9	def _handle_list(args):
9	from ttk.core_modules.testcase_manager import UniversalTestcaseFactory	10	from ttk.core_modules.testcase_manager import UniversalTestcaseFactory
10	- with open(args.input) as f:	11	+
11	- factory = UniversalTestcaseFactory(f, skip_validate=True)	12	+ factory = UniversalTestcaseFactory.from_path(args.input, args.sheet, skip_validate=True)
12	cases = factory.testcases	13	cases = factory.testcases
13	if args.operator:	14	if args.operator:
14	op_filter = args.operator.split(",")	15	op_filter = args.operator.split(",")
15	- cases = [c for c in cases if getattr(c, 'op_name', getattr(c, 'api_name', None)) in op_filter]	16	+ cases = [c for c in cases if getattr(c, "op_name", getattr(c, "api_name", None)) in op_filter]
16	for case in cases:	17	for case in cases:
17	print(case.testcase_name)	18	print(case.testcase_name)
18	print(f"\nTotal: {len(cases)} case(s)")	19	print(f"\nTotal: {len(cases)} case(s)")

Mttk/core_modules/infra/instance_base.py+123-111

	@@ -11,7 +11,6 @@
11	Profiling Instance Base Class	11	Profiling Instance Base Class
12	"""	12	"""
13		13
14	-
15	__all__ = ["InstanceBase"]	14	__all__ = ["InstanceBase"]
16		15
17		16
	@@ -20,25 +19,25 @@ import csv
20	import io	19	import io
21	import logging	20	import logging
22	import multiprocessing	21	import multiprocessing
		22	+import os
23	import shutil	23	import shutil
24	import subprocess	24	import subprocess
25	-
26	-import numpy
27	-import os
28	import time	25	import time
29	import zipfile	26	import zipfile
30	from abc import ABCMeta, abstractmethod	27	from abc import ABCMeta, abstractmethod
31	from multiprocessing.context import BaseContext	28	from multiprocessing.context import BaseContext
32	-from typing import Optional, IO, Any, Dict, List, Tuple	29	+from typing import IO, Any, Dict, List, Optional, Tuple
		30	+
		31	+import numpy
		32	+
		33	+from ...utilities import VERSION, cpu_count, get_global_storage, list_append_union, table_print
		34	+from ..tbe_multiprocessing import SimpleCommandProcess
		35	+from ..testcase_manager import TestcaseBase, UniversalTestcaseFactory
33		36
34	# Third-Party Packages	37	# Third-Party Packages
35	from .process_group import ProcessGroup	38	from .process_group import ProcessGroup
36	-from .task import TaskA, TaskType, TaskKeeper
37	from .profile_object import ProfileObject	39	from .profile_object import ProfileObject
38	-from ..tbe_multiprocessing import SimpleCommandProcess	40	+from .task import TaskA, TaskKeeper, TaskType
39	-from ..testcase_manager import UniversalTestcaseFactory, TestcaseBase
40	-from ...utilities import get_global_storage, VERSION, table_print, list_append_union
41	-from ...utilities import cpu_count
42		41
43		42
44	class InstanceBase(metaclass=ABCMeta):	43	class InstanceBase(metaclass=ABCMeta):
	@@ -90,7 +89,7 @@ class InstanceBase(metaclass=ABCMeta):
90	@staticmethod	89	@staticmethod
91	def _read_existing_header(path: str):	90	def _read_existing_header(path: str):
92	try:	91	try:
93	- with open(path, newline='', encoding='utf-8') as f:	92	+ with open(path, newline="", encoding="utf-8") as f:
94	reader = csv.reader(f)	93	reader = csv.reader(f)
95	return next(reader, None)	94	return next(reader, None)
96	except (UnicodeDecodeError, csv.Error, OSError):	95	except (UnicodeDecodeError, csv.Error, OSError):
	@@ -145,9 +144,11 @@ class InstanceBase(metaclass=ABCMeta):
145	self._close_idle_processes()	144	self._close_idle_processes()
146	self._summary_print(self.print_cycle)	145	self._summary_print(self.print_cycle)
147	if self.total_case_count == self.completed_case_count:	146	if self.total_case_count == self.completed_case_count:
148	- logging.info(f"ttk Profiling complete")	147	+ logging.info("ttk Profiling complete")
149	break	148	break
150	- time.sleep(0.01) # pacing before next iteration; avoids busy-spin 100% CPU on one core. Placed after the break-check so we don't sleep on the exiting iteration.	149	+ time.sleep(
		150	+ 0.01
		151	+ ) # pacing before next iteration; avoids busy-spin 100% CPU on one core. Placed after the break-check so we don't sleep on the exiting iteration.
151	# close all processes	152	# close all processes
152	self.close_subprocesses()	153	self.close_subprocesses()
153	# batch consistency post-processing (level=2)	154	# batch consistency post-processing (level=2)
	@@ -160,21 +161,17 @@ class InstanceBase(metaclass=ABCMeta):
160	# Create process for every usable device	161	# Create process for every usable device
161	if self.switches.process_per_device is None:	162	if self.switches.process_per_device is None:
162	# Use 80% of total cpu cores, not exceed 4	163	# Use 80% of total cpu cores, not exceed 4
163	- self.switches.process_per_device = min(max(int(cpu_count() * 0.8) //	164	+ self.switches.process_per_device = min(max(int(cpu_count() * 0.8) // len(self.used_device), 1), 4)
164	- len(self.used_device), 1),
165	- 4)
166	# Not exceed testcase count	165	# Not exceed testcase count
167	- self.switches.process_per_device = min(self.switches.process_per_device,	166	+ self.switches.process_per_device = min(self.switches.process_per_device, len(self.flatten_testcases))
168	- len(self.flatten_testcases))	167	+ self.used_device = self.used_device[: len(self.flatten_testcases)]
169	- self.used_device = self.used_device[:len(self.flatten_testcases)]
170	logging.info(f"Process per device: {self.switches.process_per_device}")	168	logging.info(f"Process per device: {self.switches.process_per_device}")
171	# Prepare SubProcesses	169	# Prepare SubProcesses
172	logging.info("Preparing Task Executors...")	170	logging.info("Preparing Task Executors...")
173	for dev_id in self.used_device:	171	for dev_id in self.used_device:
174	- self.process_groups[dev_id] = ProcessGroup(dev_id,	172	+ self.process_groups[dev_id] = ProcessGroup(
175	- self.switches.process_per_device,	173	+ dev_id, self.switches.process_per_device, self.mp_context, timeout=self.switches.proc_timeout
176	- self.mp_context,	174	+ )
177	- timeout=self.switches.proc_timeout)
178	os.environ["TTK_LOAD_TF"] = "1"	175	os.environ["TTK_LOAD_TF"] = "1"
179	for pg in self.process_groups.values():	176	for pg in self.process_groups.values():
180	while not pg.is_ready():	177	while not pg.is_ready():
	@@ -188,7 +185,7 @@ class InstanceBase(metaclass=ABCMeta):
188	configured endpoints; writes health state to TTK_XPU_HEALTH_PATH.	185	configured endpoints; writes health state to TTK_XPU_HEALTH_PATH.
189	"""	186	"""
190	try:	187	try:
191	- from ttk.remote import is_remote_configured, get_tenant_id	188	+ from ttk.remote import get_tenant_id, is_remote_configured
192	from ttk.remote.config import get_remote_config	189	from ttk.remote.config import get_remote_config
193	from ttk.remote.heartbeat import heartbeat_loop	190	from ttk.remote.heartbeat import heartbeat_loop
194	from ttk.remote.heartbeat_manager import HeartbeatManager	191	from ttk.remote.heartbeat_manager import HeartbeatManager
	@@ -208,6 +205,7 @@ class InstanceBase(metaclass=ABCMeta):
208	# NB: tls_from_config call is OUTSIDE the ImportError try/except above so a	205	# NB: tls_from_config call is OUTSIDE the ImportError try/except above so a
209	# cert/key mismatch raises as a loud startup failure (not swallowed).	206	# cert/key mismatch raises as a loud startup failure (not swallowed).
210	from ttk.remote.tls import tls_from_config	207	from ttk.remote.tls import tls_from_config
		208	+
211	tls = tls_from_config(config)	209	tls = tls_from_config(config)
212	self.heartbeat_manager = HeartbeatManager(	210	self.heartbeat_manager = HeartbeatManager(
213	heartbeat_target=heartbeat_loop,	211	heartbeat_target=heartbeat_loop,
	@@ -258,21 +256,20 @@ class InstanceBase(metaclass=ABCMeta):
258	if self.switches.summary_print:	256	if self.switches.summary_print:
259	self._get_head_commit_id()	257	self._get_head_commit_id()
260	try:	258	try:
261	- percentage = int(self.completed_case_count /	259	+ percentage = int(self.completed_case_count / self.total_case_count * 100)
262	- self.total_case_count * 100)	260	+ except Exception:
263	- except:
264	percentage = "?"	261	percentage = "?"
265	- title = (f"Version: {VERSION} "	262	+ title = (
266	- f"Summary (Device Total: {self.switches.device_count}) "	263	+ f"Version: {VERSION} "
267	- f"Progress: {percentage}% "	264	+ f"Summary (Device Total: {self.switches.device_count}) "
268	- f"{self.completed_case_count} / {self.total_case_count} "	265	+ f"Progress: {percentage}% "
269	- f"ET: {int(now - self.start_timestamp)}s "	266	+ f"{self.completed_case_count} / {self.total_case_count} "
270	- f"Rev: {self._commit_id}",)	267	+ f"ET: {int(now - self.start_timestamp)}s "
		268	+ f"Rev: {self._commit_id}",
		269	+ )
271		270
272	- loop_count = len(self.used_device) // 2 \	271	+ loop_count = len(self.used_device) // 2 if self.used_device else self.switches.device_count // 2
273	- if self.used_device else self.switches.device_count // 2	272	+ remain_count = len(self.used_device) % 2 if self.used_device else self.switches.device_count % 2
274	- remain_count = len(self.used_device) % 2 \
275	- if self.used_device else self.switches.device_count % 2
276	lines = [title]	273	lines = [title]
277	for loop in range(loop_count):	274	for loop in range(loop_count):
278	if self.used_device:	275	if self.used_device:
	@@ -280,22 +277,31 @@ class InstanceBase(metaclass=ABCMeta):
280	dev_id_1 = self.used_device[loop * 2 + 1]	277	dev_id_1 = self.used_device[loop * 2 + 1]
281	else:	278	else:
282	dev_id_0, dev_id_1 = loop * 2, loop * 2 + 1	279	dev_id_0, dev_id_1 = loop * 2, loop * 2 + 1
283	- lines.append((*(self.device_info(dev_id_0),	280	+ lines.append(
284	- self.process_groups[dev_id_0].info()	281	+ (
285	- if dev_id_0 in self.process_groups else ''),	282	+ *(
286	- *(self.device_info(dev_id_1),	283	+ self.device_info(dev_id_0),
287	- self.process_groups[dev_id_1].info()	284	+ self.process_groups[dev_id_0].info() if dev_id_0 in self.process_groups else "",
288	- if dev_id_1 in self.process_groups else '')	285	+ ),
289	- ))	286	+ *(
		287	+ self.device_info(dev_id_1),
		288	+ self.process_groups[dev_id_1].info() if dev_id_1 in self.process_groups else "",
		289	+ ),
		290	+ )
		291	+ )
290	if remain_count:	292	if remain_count:
291	if self.used_device:	293	if self.used_device:
292	dev_id = self.used_device[-1]	294	dev_id = self.used_device[-1]
293	else:	295	else:
294	dev_id = loop_count * 2	296	dev_id = loop_count * 2
295	- lines.append((*(self.device_info(dev_id),	297	+ lines.append(
296	- self.process_groups[dev_id].info()	298	+ (
297	- if dev_id in self.process_groups else ''),	299	+ *(
298	- ))	300	+ self.device_info(dev_id),
		301	+ self.process_groups[dev_id].info() if dev_id in self.process_groups else "",
		302	+ ),
		303	+ )
		304	+ )
299	logging.info("\n" + table_print(lines))	305	logging.info("\n" + table_print(lines))
300		306
301	def _pre_exit(self):	307	def _pre_exit(self):
	@@ -340,8 +346,7 @@ class InstanceBase(metaclass=ABCMeta):
340	for r in results:	346	for r in results:
341	status = "PASS" if r["pass"] else "FAIL"	347	status = "PASS" if r["pass"] else "FAIL"
342	names = [m["testcase"] for m in r["members"]]	348	names = [m["testcase"] for m in r["members"]]
343	- logging.info(f" [{status}] group={r['batch_consistency_id'][:32]}... "	349	+ logging.info(f" [{status}] group={r['batch_consistency_id'][:32]}... members={names}")
344	- f"members={names}")
345		350
346	def _prepare_device_locks(self):	351	def _prepare_device_locks(self):
347	# TODO: device-id not start from 0 to max-count in docker.	352	# TODO: device-id not start from 0 to max-count in docker.
	@@ -375,7 +380,7 @@ class InstanceBase(metaclass=ABCMeta):
375	valid_count += 1	380	valid_count += 1
376	else:	381	else:
377	reason = tc.fail_reason or "UNKNOWN"	382	reason = tc.fail_reason or "UNKNOWN"
378	- invalid_cases.append((tc.testcase_name, getattr(tc, 'api_name', ''), reason))	383	+ invalid_cases.append((tc.testcase_name, getattr(tc, "api_name", ""), reason))
379	reason_counts[reason] = reason_counts.get(reason, 0) + 1	384	reason_counts[reason] = reason_counts.get(reason, 0) + 1
380	invalid_count = len(invalid_cases)	385	invalid_count = len(invalid_cases)
381	total = valid_count + invalid_count	386	total = valid_count + invalid_count
	@@ -397,15 +402,14 @@ class InstanceBase(metaclass=ABCMeta):
397	logging.info("All testcases passed validation")	402	logging.info("All testcases passed validation")
398		403
399	def _write_validate_result(self, invalid_cases, valid_count, invalid_count):	404	def _write_validate_result(self, invalid_cases, valid_count, invalid_count):
400	- import pathlib
401	result_path = self.switches.output_file_name	405	result_path = self.switches.output_file_name
402	- if not result_path.endswith('.csv'):	406	+ if not result_path.endswith(".csv"):
403	- result_path += '.csv'	407	+ result_path += ".csv"
404	- with open(result_path, 'w', newline='', encoding='utf-8') as f:	408	+ with open(result_path, "w", newline="", encoding="utf-8") as f:
405	writer = csv.writer(f)	409	writer = csv.writer(f)
406	writer.writerow(("testcase_name", "api_name", "status", "fail_reason"))	410	writer.writerow(("testcase_name", "api_name", "status", "fail_reason"))
407	for tc in sorted(self.flatten_testcases, key=lambda t: t.testcase_name):	411	for tc in sorted(self.flatten_testcases, key=lambda t: t.testcase_name):
408	- api = getattr(tc, 'api_name', '')	412	+ api = getattr(tc, "api_name", "")
409	if tc.is_valid:	413	if tc.is_valid:
410	writer.writerow((tc.testcase_name, api, "VALID", ""))	414	writer.writerow((tc.testcase_name, api, "VALID", ""))
411	else:	415	else:
	@@ -413,15 +417,17 @@ class InstanceBase(metaclass=ABCMeta):
413	logging.info(f"Validate result written to {result_path}")	417	logging.info(f"Validate result written to {result_path}")
414		418
415	def _load_cases(self, testcase_file: str):	419	def _load_cases(self, testcase_file: str):
		420	+ if testcase_file.lower().endswith((".xlsx", ".xlsm")):
		421	+ self._load_case_from_table(testcase_file)
		422	+ return
416	use_csv_mode = False	423	use_csv_mode = False
417	try:	424	try:
418	self._load_case_from_zip(testcase_file)	425	self._load_case_from_zip(testcase_file)
419	except zipfile.BadZipFile:	426	except zipfile.BadZipFile:
420	use_csv_mode = True	427	use_csv_mode = True
421	if use_csv_mode:	428	if use_csv_mode:
422	- logging.info("Input testcase file is not a valid zip file, "	429	+ logging.info("Input testcase file is not a valid zip file, switch to normal csv mode...")
423	- "switch to normal csv mode...")	430	+ self._load_case_from_table(testcase_file)
424	- self._load_case_from_csv(testcase_file)
425		431
426	def _check_duplicate_case(self):	432	def _check_duplicate_case(self):
427	testcase_names = set()	433	testcase_names = set()
	@@ -441,13 +447,15 @@ class InstanceBase(metaclass=ABCMeta):
441	if not self.switches.input_files:	447	if not self.switches.input_files:
442	raise RuntimeError("Please specify input csv files !!!")	448	raise RuntimeError("Please specify input csv files !!!")
443	if not self.result_path:	449	if not self.result_path:
444	- split_input_path = self.switches.input_files[0].split(".")	450	+ root, _ = os.path.splitext(self.switches.input_files[0])
445	- split_input_path[-2] += "_result"	451	+ sheet = self._resolved_output_sheet()
446	- self.result_path = '.'.join(split_input_path)	452	+ tag = ""
447	- logging.info(f"Output csv file is not specified. "	453	+ if sheet:
448	- f"It will be set as {self.result_path}")	454	+ tag = "_" + sheet.translate(str.maketrans({c: "_" for c in '\\/:*?"<>\| '}))
449	- if not self.result_path.endswith('.csv'):	455	+ self.result_path = root + tag + "_result.csv"
450	- self.result_path += '.csv'	456	+ logging.info(f"Output csv file is not specified. It will be set as {self.result_path}")
		457	+ if not self.result_path.endswith(".csv"):
		458	+ self.result_path += ".csv"
451	parent_dir = os.path.dirname(self.result_path)	459	parent_dir = os.path.dirname(self.result_path)
452	if parent_dir and not os.path.isdir(parent_dir):	460	if parent_dir and not os.path.isdir(parent_dir):
453	os.makedirs(parent_dir, exist_ok=True)	461	os.makedirs(parent_dir, exist_ok=True)
	@@ -459,43 +467,51 @@ class InstanceBase(metaclass=ABCMeta):
459	existing_header = self._read_existing_header(self.result_path)	467	existing_header = self._read_existing_header(self.result_path)
460	header_match = existing_header is not None and tuple(existing_header) == self.case_result_titles	468	header_match = existing_header is not None and tuple(existing_header) == self.case_result_titles
461	if header_match:	469	if header_match:
462	- self.result_csv_file = open(self.result_path, newline='', mode='a+')	470	+ self.result_csv_file = open(self.result_path, newline="", mode="a+")
463	self.result_csv_writer = csv.writer(self.result_csv_file)	471	self.result_csv_writer = csv.writer(self.result_csv_file)
464	self._header_flushed = True	472	self._header_flushed = True
465	self._precision_status_idx = self._resolve_precision_status_idx(self.case_result_titles)	473	self._precision_status_idx = self._resolve_precision_status_idx(self.case_result_titles)
466	logging.info(f"Append mode: appending to existing {self.result_path}")	474	logging.info(f"Append mode: appending to existing {self.result_path}")
467	else:	475	else:
468	- logging.warning(f"Append mode: existing file header does not match "	476	+ logging.warning(
469	- f"(file has {len(existing_header) if existing_header else 0} columns, "	477	+ f"Append mode: existing file header does not match "
470	- f"current expects {len(self.case_result_titles)}). "	478	+ f"(file has {len(existing_header) if existing_header else 0} columns, "
471	- f"Overwriting {self.result_path}")	479	+ f"current expects {len(self.case_result_titles)}). "
472	- self.result_csv_file = open(self.result_path, newline='', mode='w+')	480	+ f"Overwriting {self.result_path}"
		481	+ )
		482	+ self.result_csv_file = open(self.result_path, newline="", mode="w+")
473	self.result_csv_writer = csv.writer(self.result_csv_file)	483	self.result_csv_writer = csv.writer(self.result_csv_file)
474	self._flush(self.case_result_titles)	484	self._flush(self.case_result_titles)
475	else:	485	else:
476	- self.result_csv_file = open(self.result_path, newline='', mode='w+')	486	+ self.result_csv_file = open(self.result_path, newline="", mode="w+")
477	self.result_csv_writer = csv.writer(self.result_csv_file)	487	self.result_csv_writer = csv.writer(self.result_csv_file)
478	self._prepare_output_titles()	488	self._prepare_output_titles()
479	self._flush(self.case_result_titles)	489	self._flush(self.case_result_titles)
480		490
		491	+ def _resolved_output_sheet(self):
		492	+ """Worksheet name to embed in the default output name (xlsx only)."""
		493	+ src = self.switches.input_files[0] if self.switches.input_files else ""
		494	+ if not src.lower().endswith((".xlsx", ".xlsm")):
		495	+ return None
		496	+ from ...utilities import resolved_sheet
		497	+
		498	+ return resolved_sheet(src, getattr(self.switches, "sheet", None))
		499	+
481	def _prepare_output_titles(self):	500	def _prepare_output_titles(self):
482	first_testcase = next(iter(self.flatten_testcases))	501	first_testcase = next(iter(self.flatten_testcases))
483	- self.case_result_titles = self.profile_object.output_titles(first_testcase,	502	+ self.case_result_titles = self.profile_object.output_titles(first_testcase, self.case_original_headers)
484	- self.case_original_headers)
485		503
486	def _initialize_device_lock(self) -> tuple:	504	def _initialize_device_lock(self) -> tuple:
487	self.mp_manager = self.mp_context.Manager()	505	self.mp_manager = self.mp_context.Manager()
488	if self.switches.device_whitelist:	506	if self.switches.device_whitelist:
489	- blacklist = list(set([i for i in range(self.switches.device_count)]) -	507	+ blacklist = list(set([i for i in range(self.switches.device_count)]) - set(self.switches.device_whitelist))
490	- set(self.switches.device_whitelist))	508	+ self.switches.device_blacklist = list(set(blacklist).union(set(self.switches.device_blacklist)))
491	- self.switches.device_blacklist = list(set(blacklist).union(
492	- set(self.switches.device_blacklist)))
493	# Print Device blacklist info	509	# Print Device blacklist info
494	if self.switches.device_blacklist:	510	if self.switches.device_blacklist:
495	- logging.info(f"Device {self.switches.device_blacklist} "	511	+ logging.info(f"Device {self.switches.device_blacklist} has been blacklisted, removing...")
496	- f"has been blacklisted, removing...")	512	+ available = tuple(
497	- available = tuple(True if n not in self.switches.device_blacklist else None	513	+ True if n not in self.switches.device_blacklist else None for n in range(self.switches.device_count)
498	- for n in range(self.switches.device_count))	514	+ )
499	available_devices = [i for i, v in enumerate(available) if v is not None]	515	available_devices = [i for i, v in enumerate(available) if v is not None]
500	SimpleCommandProcess.initialize_device_locks(available_devices, self.mp_manager)	516	SimpleCommandProcess.initialize_device_locks(available_devices, self.mp_manager)
501	return available	517	return available
	@@ -524,11 +540,13 @@ class InstanceBase(metaclass=ABCMeta):
524	def _output_progress(self):	540	def _output_progress(self):
525	if self.switches.progress_output:	541	if self.switches.progress_output:
526	now = time.time()	542	now = time.time()
527	- with open(self.switches.progress_output, 'w') as f:	543	+ with open(self.switches.progress_output, "w") as f:
528	- f.write(f"StartAt:{str(self.start_timestamp)}\n"	544	+ f.write(
529	- f"ElapsedTime:{str(now - self.start_timestamp)}\n"	545	+ f"StartAt:{str(self.start_timestamp)}\n"
530	- f"TotalCases:{str(self.total_case_count)}\n"	546	+ f"ElapsedTime:{str(now - self.start_timestamp)}\n"
531	- f"CompletedCases:{str(self.completed_case_count)}")	547	+ f"TotalCases:{str(self.total_case_count)}\n"
		548	+ f"CompletedCases:{str(self.completed_case_count)}"
		549	+ )
532		550
533	def _push_at_least_one_prof_task_to_process(self):	551	def _push_at_least_one_prof_task_to_process(self):
534	if self._multi_device_running:	552	if self._multi_device_running:
	@@ -597,22 +615,17 @@ class InstanceBase(metaclass=ABCMeta):
597	self.testcase_complete()	615	self.testcase_complete()
598		616
599	def _handle_multi_device_task_result(self, task: TaskA, proc: SimpleCommandProcess):	617	def _handle_multi_device_task_result(self, task: TaskA, proc: SimpleCommandProcess):
600	- case_name = task.testcase.testcase_name
601	result = proc.get_result()	618	result = proc.get_result()
602	pid = proc.get_pid()	619	pid = proc.get_pid()
603	- dev_id = None	620	+ for pg in self.process_groups.values():
604	- for pg_dev_id, pg in self.process_groups.items():
605	if proc in pg.process_to_task:	621	if proc in pg.process_to_task:
606	- dev_id = pg_dev_id
607	break	622	break
608	if isinstance(result, (SystemError, RuntimeError)):	623	if isinstance(result, (SystemError, RuntimeError)):
609	proc.resurrect()	624	proc.resurrect()
610	if isinstance(result, SystemError):	625	if isinstance(result, SystemError):
611	- output_data = self.profile_object.handle_task_result_system_error(	626	+ output_data = self.profile_object.handle_task_result_system_error(task, result, "MultiDevice", pid)
612	- task, result, "MultiDevice", pid)
613	else:	627	else:
614	- output_data = self.profile_object.handle_task_result_runtime_error(	628	+ output_data = self.profile_object.handle_task_result_runtime_error(task, result, pid)
615	- task, result, pid)
616	elif result is not None:	629	elif result is not None:
617	output_data, kill_proc = self.profile_object.handle_task_result_complete(task, result)	630	output_data, kill_proc = self.profile_object.handle_task_result_complete(task, result)
618	if self.switches.proc_no_reuse or kill_proc or task.is_multi_device():	631	if self.switches.proc_no_reuse or kill_proc or task.is_multi_device():
	@@ -631,9 +644,7 @@ class InstanceBase(metaclass=ABCMeta):
631	output_data = self.profile_object.handle_task_result_none(task)	644	output_data = self.profile_object.handle_task_result_none(task)
632	elif isinstance(result, SystemError):	645	elif isinstance(result, SystemError):
633	proc.resurrect()	646	proc.resurrect()
634	- output_data = self.profile_object.handle_task_result_system_error(task, result,	647	+ output_data = self.profile_object.handle_task_result_system_error(task, result, proc.current_stage(), pid)
635	- proc.current_stage(),
636	- pid)
637	elif isinstance(result, RuntimeError):	648	elif isinstance(result, RuntimeError):
638	proc.resurrect()	649	proc.resurrect()
639	output_data = self.profile_object.handle_task_result_runtime_error(task, result, pid)	650	output_data = self.profile_object.handle_task_result_runtime_error(task, result, pid)
	@@ -688,32 +699,33 @@ class InstanceBase(metaclass=ABCMeta):
688	logging.info(f"Reading zipped testcase file {file.filename} ...")	699	logging.info(f"Reading zipped testcase file {file.filename} ...")
689	with zipped_file.open(file) as real_file:	700	with zipped_file.open(file) as real_file:
690	testcase_manager = UniversalTestcaseFactory(	701	testcase_manager = UniversalTestcaseFactory(
691	- io.TextIOWrapper(real_file, encoding="UTF-8", newline=''))	702	+ io.TextIOWrapper(real_file, encoding="UTF-8", newline="")
		703	+ )
692	self.flatten_testcases.extend(testcase_manager.get())	704	self.flatten_testcases.extend(testcase_manager.get())
693	list_append_union(self.case_original_headers, testcase_manager.header)	705	list_append_union(self.case_original_headers, testcase_manager.header)
694		706
695	- def _load_case_from_csv(self, testcase_path: str):	707	+ def _load_case_from_table(self, testcase_path: str):
696	- logging.info("Reading normal csv testcases...")	708	+ sheet = getattr(self.switches, "sheet", None)
697	- with open(testcase_path, newline='', encoding='utf-8') as file:	709	+ logging.info(f"Reading testcases from {testcase_path}{f' sheet={sheet}' if sheet else ''} ...")
698	- testcase_manager = UniversalTestcaseFactory(file)	710	+ testcase_manager = UniversalTestcaseFactory.from_path(testcase_path, sheet)
699	- self.flatten_testcases.extend(testcase_manager.get())	711	+ self.flatten_testcases.extend(testcase_manager.get())
700	- list_append_union(self.case_original_headers, testcase_manager.header)	712	+ list_append_union(self.case_original_headers, testcase_manager.header)
701		713
702	def _get_head_commit_id(self):	714	def _get_head_commit_id(self):
703	if self._commit_id is not None:	715	if self._commit_id is not None:
704	return	716	return
705	- if shutil.which('git') is None:	717	+ if shutil.which("git") is None:
706	self._commit_id = "UNKNOWN"	718	self._commit_id = "UNKNOWN"
707	return	719	return
708		720
709	try:	721	try:
710	- result = subprocess.run(["git", "rev-parse", "--short", "HEAD"],	722	+ result = subprocess.run(
711	- capture_output=True, text=True,	723	+ ["git", "rev-parse", "--short", "HEAD"], capture_output=True, text=True, shell=False, check=False
712	- shell=False, check=False)	724	+ )
713	if result.returncode != 0:	725	if result.returncode != 0:
714	self._commit_id = "UNKNOWN"	726	self._commit_id = "UNKNOWN"
715	else:	727	else:
716	- self._commit_id = result.stdout.split('\n')[0]	728	+ self._commit_id = result.stdout.split("\n")[0]
717	finally:	729	finally:
718	if self._commit_id is None:	730	if self._commit_id is None:
719	self._commit_id = "UNKNOWN"	731	self._commit_id = "UNKNOWN"

Mttk/core_modules/testcase_manager/testcase_aclnn.py+15-24

	@@ -17,21 +17,15 @@ __all__ = ["TestcaseAclnn", "AclnnParamPlan"]
17	# Standard Packages	17	# Standard Packages
18	import copy	18	import copy
19	import logging	19	import logging
20	-import numpy.random	20	+from typing import Any, Dict, List, Optional, Tuple, Union
21	-from typing import Dict, List, Optional, Tuple, Union, Any
22		21
23	-try:	22	+import numpy.random
24	- from collections.abc import Callable
25	-except ImportError:
26	- from collections import Callable
27		23
28	# Third-party Packages	24	# Third-party Packages
29	-from .testcase_tensor_api_base import TensorApiTestcaseBase	25	+from ...utilities import get, get_dtype_width, parse_dtype, shape_product, shape_product_with_strides, shape_stride
		26	+from ..aclnn import OpApiInfo, OpApiInfoKeeper
30	from .field_types import FIELD_TYPES	27	from .field_types import FIELD_TYPES
31	-from ..aclnn import OpApiInfoKeeper, OpApiInfo	28	+from .testcase_tensor_api_base import TensorApiTestcaseBase
32	-from ...utilities import get, shape_stride, shape_product_with_strides
33	-from ...utilities import shape_product, parse_dtype, get_dtype_width
34	-from ...utilities.container_utils import infer_list_distribution_from_nesting
35		29
36		30
37	class AclnnParamPlan:	31	class AclnnParamPlan:
	@@ -94,7 +88,7 @@ class AclnnParamPlan:
94	tensor_queue = list(tensors)	88	tensor_queue = list(tensors)
95	scalar_queue = list(scalars)	89	scalar_queue = list(scalars)
96	param_names = set()	90	param_names = set()
97	- for kind, name, acl_type, default in self.param_layout:	91	+ for kind, name, _acl_type, default in self.param_layout:
98	param_names.add(name)	92	param_names.add(name)
99	if kind == self.TENSOR:	93	if kind == self.TENSOR:
100	args.append(tensor_queue.pop(0))	94	args.append(tensor_queue.pop(0))
	@@ -285,7 +279,7 @@ class TestcaseAclnn(TensorApiTestcaseBase):
285	bytes_lst.append(	279	bytes_lst.append(
286	shape_product(self.flat_storage_shape(idx)) * get_dtype_width(get(self.flat_tensor_dtypes, idx))	280	shape_product(self.flat_storage_shape(idx)) * get_dtype_width(get(self.flat_tensor_dtypes, idx))
287	)	281	)
288	- except:	282	+ except Exception:
289	bytes_lst.append(0)	283	bytes_lst.append(0)
290	return sum(bytes_lst)	284	return sum(bytes_lst)
291		285
	@@ -457,6 +451,7 @@ class TestcaseAclnn(TensorApiTestcaseBase):
457	"""Return attrs for XPU dispatch: pure_attrs + scalars (scalars are not in	451	"""Return attrs for XPU dispatch: pure_attrs + scalars (scalars are not in
458	X-Input-Schema, which only transports tensors, so they must go via attrs)."""	452	X-Input-Schema, which only transports tensors, so they must go via attrs)."""
459	from ...utilities.container_utils import deep_flatten	453	from ...utilities.container_utils import deep_flatten
		454	+
460	attrs = dict(self.pure_attrs)	455	attrs = dict(self.pure_attrs)
461	op_api_info = OpApiInfoKeeper().info_of(self.api_name)	456	op_api_info = OpApiInfoKeeper().info_of(self.api_name)
462	if op_api_info and self.scalars is not None:	457	if op_api_info and self.scalars is not None:
	@@ -465,7 +460,7 @@ class TestcaseAclnn(TensorApiTestcaseBase):
465	if idx < len(flat_scalars):	460	if idx < len(flat_scalars):
466	s = flat_scalars[idx]	461	s = flat_scalars[idx]
467	if s is not None:	462	if s is not None:
468	- attrs[name] = s.item() if hasattr(s, 'item') else s	463	+ attrs[name] = s.item() if hasattr(s, "item") else s
469	return attrs	464	return attrs
470		465
471	@property	466	@property
	@@ -516,7 +511,7 @@ class TestcaseAclnn(TensorApiTestcaseBase):
516	if self.device_ids is not None and isinstance(self.device_ids, str):	511	if self.device_ids is not None and isinstance(self.device_ids, str):
517	raw = self.device_ids.strip()	512	raw = self.device_ids.strip()
518	if raw:	513	if raw:
519	- self.device_ids = tuple(int(d.strip()) for d in raw.split(',') if d.strip())	514	+ self.device_ids = tuple(int(d.strip()) for d in raw.split(",") if d.strip())
520	else:	515	else:
521	self.device_ids = None	516	self.device_ids = None
522	if self.device_ids is not None and not isinstance(self.device_ids, tuple):	517	if self.device_ids is not None and not isinstance(self.device_ids, tuple):
	@@ -640,12 +635,12 @@ class TestcaseAclnn(TensorApiTestcaseBase):
640	return self.device_ids is not None and len(self.device_ids) > 1	635	return self.device_ids is not None and len(self.device_ids) > 1
641		636
642	def parse_device_ids(self, raw_value):	637	def parse_device_ids(self, raw_value):
643	- if raw_value is None or raw_value == '':	638	+ if raw_value is None or raw_value == "":
644	self.device_ids = None	639	self.device_ids = None
645	self.my_rank = None	640	self.my_rank = None
646	return	641	return
647	if isinstance(raw_value, str):	642	if isinstance(raw_value, str):
648	- self.device_ids = tuple(int(d.strip()) for d in raw_value.split(',') if d.strip())	643	+ self.device_ids = tuple(int(d.strip()) for d in raw_value.split(",") if d.strip())
649	elif isinstance(raw_value, (tuple, list)):	644	elif isinstance(raw_value, (tuple, list)):
650	self.device_ids = tuple(int(d) for d in raw_value)	645	self.device_ids = tuple(int(d) for d in raw_value)
651	else:	646	else:
	@@ -690,7 +685,7 @@ class TestcaseAclnn(TensorApiTestcaseBase):
690	if len(view_shape) != len(view_stride):	685	if len(view_shape) != len(view_stride):
691	logging.error(f"Rank of view_shape/view_strides should be same: {view_shape} vs {view_stride}")	686	logging.error(f"Rank of view_shape/view_strides should be same: {view_shape} vs {view_stride}")
692	return False	687	return False
693	- for idx, v in enumerate(view_stride):	688	+ for _, v in enumerate(view_stride):
694	if v < 0:	689	if v < 0:
695	logging.error(f"Negative view_strides are not supported: {view_stride}.")	690	logging.error(f"Negative view_strides are not supported: {view_stride}.")
696	return False	691	return False
	@@ -881,7 +876,7 @@ class TestcaseAclnn(TensorApiTestcaseBase):
881	try:	876	try:
882	if self.tensor_dtypes:	877	if self.tensor_dtypes:
883	self.tensor_dtypes = self._recursively_parse(self.tensor_dtypes, parse_dtype)	878	self.tensor_dtypes = self._recursively_parse(self.tensor_dtypes, parse_dtype)
884	- except:	879	+ except Exception:
885	self.is_valid = False	880	self.is_valid = False
886	self.fail_reason = "TENSOR_DTYPES_INVALID"	881	self.fail_reason = "TENSOR_DTYPES_INVALID"
887	logging.exception(f"Tensor dtypes parse failed: {self.tensor_dtypes}")	882	logging.exception(f"Tensor dtypes parse failed: {self.tensor_dtypes}")
	@@ -893,7 +888,7 @@ class TestcaseAclnn(TensorApiTestcaseBase):
893	try:	888	try:
894	if self.scalar_dtypes:	889	if self.scalar_dtypes:
895	self.scalar_dtypes = self._recursively_parse(self.scalar_dtypes, parse_dtype)	890	self.scalar_dtypes = self._recursively_parse(self.scalar_dtypes, parse_dtype)
896	- except:	891	+ except Exception:
897	self.is_valid = False	892	self.is_valid = False
898	self.fail_reason = "SCALAR_DTYPES_INVALID"	893	self.fail_reason = "SCALAR_DTYPES_INVALID"
899	logging.exception(f"Scalar dtypes parse failed: {self.scalar_dtypes}")	894	logging.exception(f"Scalar dtypes parse failed: {self.scalar_dtypes}")
	@@ -1057,9 +1052,6 @@ class TestcaseAclnn(TensorApiTestcaseBase):
1057	inplace_indices = []	1052	inplace_indices = []
1058	for idx, element in enumerate(self.tensor_view_shapes):	1053	for idx, element in enumerate(self.tensor_view_shapes):
1059	param_name = op_api_info.tensors[idx]	1054	param_name = op_api_info.tensors[idx]
1060	- is_nested = (
1061	- isinstance(element, (tuple, list)) and len(element) > 0 and isinstance(element[0], (tuple, list))
1062	- )
1063	if param_name in ref_lst:	1055	if param_name in ref_lst:
1064	if element is None:	1056	if element is None:
1065	logging.info(f"Inplace parameter [{param_name}] is None (nullptr), skipping.")	1057	logging.info(f"Inplace parameter [{param_name}] is None (nullptr), skipping.")
	@@ -1074,7 +1066,6 @@ class TestcaseAclnn(TensorApiTestcaseBase):
1074	"gradOut",	1066	"gradOut",
1075	"grad_output",	1067	"grad_output",
1076	"attentionOut",	1068	"attentionOut",
1077	- "dOut",
1078	}	1069	}
1079	)	1070	)
1080		1071

Mttk/core_modules/testcase_manager/testcase_manager.py+31-26

	@@ -15,12 +15,12 @@ __all__ = ["UniversalTestcaseFactory"]
15		15
16		16
17	# Standard Packages	17	# Standard Packages
18	-import csv
19	import logging	18	import logging
20	import random	19	import random
21	from typing import Any, Dict, List, Optional, Set, TextIO	20	from typing import Any, Dict, List, Optional, Set, TextIO
22		21
23	from ...utilities import get_global_storage, set_process_name, set_thread_name	22	from ...utilities import get_global_storage, set_process_name, set_thread_name
		23	+from ...utilities.table_reader import read_csv_rows, read_table
24		24
25	# Third-Party Packages	25	# Third-Party Packages
26	from .testcase_base import TestcaseBase	26	from .testcase_base import TestcaseBase
	@@ -45,6 +45,23 @@ class UniversalTestcaseFactory:
45	"""	45	"""
46	Store the whole Testcases in the csv file into memory	46	Store the whole Testcases in the csv file into memory
47	"""	47	"""
		48	+ self._init_common(skip_validate)
		49	+ header, rows = read_csv_rows(file)
		50	+ self._init_from_rows(header, rows)
		51	+
		52	+ @classmethod
		53	+ def from_path(cls, path: str, sheet: Optional[str] = None, skip_validate=False):
		54	+ """
		55	+ Load testcases from a CSV or XLSX file by path. XLSX uses openpyxl
		56	+ and honors ``sheet`` (default: first worksheet); CSV ignores it.
		57	+ """
		58	+ self = cls.__new__(cls)
		59	+ self._init_common(skip_validate)
		60	+ header, rows = read_table(path, sheet)
		61	+ self._init_from_rows(header, rows)
		62	+ return self
		63	+
		64	+ def _init_common(self, skip_validate=False):
48	# Raw rows	65	# Raw rows
49	self.raw_data: List[List[str]] = []	66	self.raw_data: List[List[str]] = []
50	# Headers	67	# Headers
	@@ -59,8 +76,10 @@ class UniversalTestcaseFactory:
59		76
60	set_process_name("TestcaseManager")	77	set_process_name("TestcaseManager")
61	set_thread_name("Initialization")	78	set_thread_name("Initialization")
62	- # read csv file.	79	+
63	- self._read_csv(file)	80	+ def _init_from_rows(self, header: List[str], rows: List[List[str]]):
		81	+ self.header = list(header)
		82	+ self.raw_data = [list(row) for row in rows]
64	self._testcase_hdr_check()	83	self._testcase_hdr_check()
65	self._parse_testcase()	84	self._parse_testcase()
66	set_process_name()	85	set_process_name()
	@@ -81,8 +100,8 @@ class UniversalTestcaseFactory:
81	set_thread_name(testcase_struct.testcase_name)	100	set_thread_name(testcase_struct.testcase_name)
82	try:	101	try:
83	testcase_struct.validate()	102	testcase_struct.validate()
84	- except:	103	+ except Exception as err:
85	- raise RuntimeError(f"Failed parsing testcase {testcase_struct.testcase_name}")	104	+ raise RuntimeError(f"Failed parsing testcase {testcase_struct.testcase_name}") from err
86		105
87	@staticmethod	106	@staticmethod
88	def set_case_default_value(testcases: List[TestcaseBase]) -> None:	107	def set_case_default_value(testcases: List[TestcaseBase]) -> None:
	@@ -165,7 +184,7 @@ class UniversalTestcaseFactory:
165	original_result = testcase_struct.original_dict[title]	184	original_result = testcase_struct.original_dict[title]
166	try:	185	try:
167	float(original_result)	186	float(original_result)
168	- except:	187	+ except Exception:
169	if original_result not in ("PASS",):	188	if original_result not in ("PASS",):
170	return True	189	return True
171	else:	190	else:
	@@ -177,7 +196,7 @@ class UniversalTestcaseFactory:
177	@staticmethod	196	@staticmethod
178	def _check_testcase_enabled(testcase_struct: TestcaseBase) -> bool:	197	def _check_testcase_enabled(testcase_struct: TestcaseBase) -> bool:
179	if not testcase_struct.is_enabled:	198	if not testcase_struct.is_enabled:
180	- logging.debug("Testcase %s skipped bcz it's disabled" % testcase_struct.testcase_name)	199	+ logging.debug(f"Testcase {testcase_struct.testcase_name} skipped bcz it's disabled")
181	return False	200	return False
182	# Skip testcase if it is disabled in current soc	201	# Skip testcase if it is disabled in current soc
183	current_soc = get_global_storage().short_soc_version	202	current_soc = get_global_storage().short_soc_version
	@@ -253,20 +272,6 @@ class UniversalTestcaseFactory:
253	logging.warning(f"Detected duplicate testcase name: {ori_name}. Rename it to {new_name}")	272	logging.warning(f"Detected duplicate testcase name: {ori_name}. Rename it to {new_name}")
254	return new_name	273	return new_name
255		274
256	- def _read_csv(self, file: TextIO):
257	- csv_reader = csv.reader(file)
258	- for row in csv_reader:
259	- row = [column.strip() for column in row]
260	- if row:
261	- self.raw_data.append(row)
262	- # First line should always be the title
263	- try:
264	- self.header = self.raw_data[0]
265	- except IndexError:
266	- logging.error("Testcase initialization received IndexError, csv file might be empty!")
267	- raise
268	- del self.raw_data[0]
269	-
270	def _testcase_hdr_check(self):	275	def _testcase_hdr_check(self):
271	set_thread_name("HeaderCheckTestcaseName")	276	set_thread_name("HeaderCheckTestcaseName")
272	# Testcase name generation	277	# Testcase name generation
	@@ -276,7 +281,7 @@ class UniversalTestcaseFactory:
276	)	281	)
277	self.header.append("testcase_name")	282	self.header.append("testcase_name")
278	for idx, row in enumerate(self.raw_data):	283	for idx, row in enumerate(self.raw_data):
279	- row.append("auto_testcase_name_%d" % (idx + 1))	284	+ row.append(f"auto_testcase_name_{idx + 1}")
280		285
281	if "api_name" in self.header:	286	if "api_name" in self.header:
282	# Auto-detect from api_name values: aclnnXxx -> aclnn, others -> framework-api	287	# Auto-detect from api_name values: aclnnXxx -> aclnn, others -> framework-api
	@@ -317,8 +322,8 @@ class UniversalTestcaseFactory:
317	header_check_set = set()	322	header_check_set = set()
318	for actual_header in self.header:	323	for actual_header in self.header:
319	if actual_header in header_check_set and actual_header not in ignored_headers:	324	if actual_header in header_check_set and actual_header not in ignored_headers:
320	- logging.error("Detected duplicate header: %s" % actual_header)	325	+ logging.error(f"Detected duplicate header: {actual_header}")
321	- raise RuntimeError("Detected duplicate header: %s" % actual_header)	326	+ raise RuntimeError(f"Detected duplicate header: {actual_header}")
322	header_check_set.add(actual_header)	327	header_check_set.add(actual_header)
323		328
324	def _parse_testcase(self):	329	def _parse_testcase(self):
	@@ -489,10 +494,10 @@ class UniversalTestcaseFactory:
489	)	494	)
490	testcase_names.add(testcase_struct.testcase_name)	495	testcase_names.add(testcase_struct.testcase_name)
491	else:	496	else:
492	- logging.warning("Duplicate testcase: %s" % testcase_struct.testcase_name)	497	+ logging.warning(f"Duplicate testcase: {testcase_struct.testcase_name}")
493	# For testcase_count selector	498	# For testcase_count selector
494	if 0 < get_global_storage().selected_testcase_count < len(self.testcases):	499	if 0 < get_global_storage().selected_testcase_count < len(self.testcases):
495	- logging.info("Selecting %d cases from all testcases" % get_global_storage().selected_testcase_count)	500	+ logging.info(f"Selecting {get_global_storage().selected_testcase_count} cases from all testcases")
496	all_indexes = random.sample(	501	all_indexes = random.sample(
497	tuple(range(len(self.testcases))), k=get_global_storage().selected_testcase_count	502	tuple(range(len(self.testcases))), k=get_global_storage().selected_testcase_count
498	)	503	)

Mttk/utilities/__init__.py+11-8

	@@ -10,18 +10,21 @@
10	"""	10	"""
11	Precious Utility Functions	11	Precious Utility Functions
12	"""	12	"""
13	-from .platform import *	13	+
14	from .classes import *	14	from .classes import *
15	-from .string_utils import *
16	from .container_utils import *	15	from .container_utils import *
17	-from .math import *
18	-from .format_utils import *
19	-from .file_utils import *
20	-from .singleton import Singleton
21	-from .dtypes import *
22	from .data import RandomData	16	from .data import RandomData
		17	+from .dtypes import *
		18	+from .file_utils import *
		19	+from .format_utils import *
		20	+from .func_dispatch import UnknownParamError, bind_by_name, framework_of, resolve_callable_str
		21	+from .math import *
		22	+from .platform import *
23	from .plog_utils import extract_plog_errors	23	from .plog_utils import extract_plog_errors
24	from .proc import *	24	from .proc import *
25	-from .func_dispatch import framework_of, bind_by_name, resolve_callable_str, UnknownParamError	25	+from .singleton import Singleton
		26	+from .string_utils import *
		27	+from .table_reader import *
		atomgit-bot8月26日 🔴 Critical changed line：本 diff 在 ttk/utilities/init.py 新增 `from .table_reader import *`（line 27），并在 testcase_manager.py:23 新增 `from ...utilities.table_reader import read_csv_rows, read_table`、bridge.py:344 新增 `from ttk.utilities.table_reader import read_table`、instance_base.py:496 新增 `from ...utilities import resolved_sheet`。 affected behavior/contract：这些导入依赖一个名为 ttk/utilities/table_reader 的新模块，但本次 PR 的 10 个变更文件中没有任何一个新增该模块（变更文件清单里没有 table_reader.py），且工作区中 `ttk/utilities/` 目录及全仓库均不存在该文件（已用 list_directory/glob 确认，grep 只能找到导入点）。 failure mode：任何入口（list/run）导入 ttk.utilities 或 ttk.core_modules.testcase_manager 时都会立即抛出 ModuleNotFoundError: No module named 'ttk.utilities.table_reader'，整个 ttk CLI 启动即失败，本 PR 的 Excel 支持功能完全不可达；属于构建/启动级故障。 suggested fix：在 PR 中补充 ttk/utilities/table_reader.py，实现 `read_csv_rows(file) -> (header, rows)`、`read_table(path, sheet=None) -> (header, rows)`（按 .csv/.xlsx/.xlsm 分派）以及 `resolved_sheet(path, sheet) -> str`；注意被删除的 _read_csv 会对每个单元格做 `column.strip()`，read_csv_rows 需保留该行为以免 CSV 值带首尾空白导致后续比较（如 api_name/框架检测、header 匹配）回归。建议：在 ttk/utilities/ 下新增 table_reader.py，提供 read_csv_rows / read_table / resolved_sheet 三个函数（CSV 用 csv 模块、xlsx/xlsm 用 openpyxl，sheet 参数缺失时取第一个工作表），并在 read_csv_rows 中保留旧 _read_csv 的逐单元格 strip 行为；否则删除这些导入点并回退到原有 CSV 路径。不准确？
		28	+
26	VERSION = "3.0.0"	29	VERSION = "3.0.0"
27	FAQ = ""	30	FAQ = ""

Mttk/utilities/classes.py+20-19

	@@ -19,10 +19,13 @@ import os
19	import pathlib	19	import pathlib
20	from dataclasses import dataclass, field	20	from dataclasses import dataclass, field
21	from enum import Enum, auto	21	from enum import Enum, auto
22	-from typing import Dict, List, Optional, Tuple, Union	22	+from typing import TYPE_CHECKING, Dict, List, Optional, Tuple, Union
23		23
24	import numpy	24	import numpy
25		25
		26	+if TYPE_CHECKING:
		27	+ from ..core_modules.testcase_manager.testcase_op import TestcaseOp
		28	+
26	# Third-party Packages	29	# Third-party Packages
27		30
28		31
	@@ -104,6 +107,7 @@ class SWITCHES:
104	"root_path",	107	"root_path",
105	"mode",	108	"mode",
106	"input_files",	109	"input_files",
		110	+ "sheet",
107	"output_file_name",	111	"output_file_name",
108	"append_mode",	112	"append_mode",
109	"logging_to_file",	113	"logging_to_file",
	@@ -203,7 +207,7 @@ class SWITCHES:
203	hw_info = get_npu_hw_info(self.dev_plat)	207	hw_info = get_npu_hw_info(self.dev_plat)
204	support_bf16 = hw_info.get("support_bf16", False)	208	support_bf16 = hw_info.get("support_bf16", False)
205	return 1 if support_bf16 else 0	209	return 1 if support_bf16 else 0
206	- except:	210	+ except Exception:
207	return 1	211	return 1
208		212
209	@property	213	@property
	@@ -243,6 +247,7 @@ class SWITCHES:
243	self.root_path = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))	247	self.root_path = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
244	self.mode: MODE = MODE.ASCEND_ONBOARD	248	self.mode: MODE = MODE.ASCEND_ONBOARD
245	self.input_files: Optional[List[str]] = None	249	self.input_files: Optional[List[str]] = None
		250	+ self.sheet: Optional[str] = None
246	self.output_file_name: Optional[str] = None	251	self.output_file_name: Optional[str] = None
247	self.append_mode: bool = False	252	self.append_mode: bool = False
248	self.logging_to_file: bool = False	253	self.logging_to_file: bool = False
	@@ -342,15 +347,11 @@ class OPTestSwitch:
342	self.prof = profiling	347	self.prof = profiling
343		348
344	def __str__(self):	349	def __str__(self):
345	- return "%s: %s, %s, %s" % (	350	+ return (
346	- self.name,	351	+ f"{self.name}: "
347	- "ENABLED" if self.enabled else "DISABLED",	352	+ f"{'ENABLED' if self.enabled else 'DISABLED'}, "
348	- "MANUAL_COMPILE"	353	+ f"{'MANUAL_COMPILE' if not self.realtime else 'RELEASE' if self.realtime == self.REUSE_BINARY_RELEASE_KERNEL else 'TE_COMPILE'}, "
349	- if not self.realtime	354	+ f"{'ONLINE' if self.prof else 'OFFLINE'}"
350	- else "RELEASE"
351	- if self.realtime == self.REUSE_BINARY_RELEASE_KERNEL
352	- else "TE_COMPILE",
353	- "ONLINE" if self.prof else "OFFLINE",
354	)	355	)
355		356
356	def use_release_bin(self):	357	def use_release_bin(self):
	@@ -438,8 +439,8 @@ class KernelJsonInfo:
438	# noinspection PyBroadException	439	# noinspection PyBroadException
439	try:	440	try:
440	json_data = json.loads(raw_json_data)	441	json_data = json.loads(raw_json_data)
441	- except Exception:	442	+ except Exception as err:
442	- raise RuntimeError("Json read failure, received json:\n%s" % raw_json_data)	443	+ raise RuntimeError(f"Json read failure, received json:\n{raw_json_data}") from err
443	return cls.from_dict(json_data)	444	return cls.from_dict(json_data)
444		445
445	@classmethod	446	@classmethod
	@@ -598,7 +599,7 @@ class BaseCompilationResult:
598	json_parsed["kernel_dir"] = self.kernel_dir	599	json_parsed["kernel_dir"] = self.kernel_dir
599	return json_parsed	600	return json_parsed
600		601
601	- def apply(self, testcase: "ttk.TestcaseOp"):	602	+ def apply(self, testcase: "TestcaseOp"):
602	pass	603	pass
603		604
604	def printf_enabled(self) -> bool:	605	def printf_enabled(self) -> bool:
	@@ -713,7 +714,7 @@ class DynamicCompilationResult(BaseCompilationResult):
713	"""Get standard value"""	714	"""Get standard value"""
714	return (self.compile_info, self.tiling_op_type, *self.base_standard_get())	715	return (self.compile_info, self.tiling_op_type, *self.base_standard_get())
715		716
716	- def apply(self, testcase: "ttk.TestcaseOp"):	717	+ def apply(self, testcase: "TestcaseOp"):
717	"""Apply dynamic result to testcase"""	718	"""Apply dynamic result to testcase"""
718	if testcase.dyn_func_params is None:	719	if testcase.dyn_func_params is None:
719	testcase.dyn_func_params = self.func_params	720	testcase.dyn_func_params = self.func_params
	@@ -789,20 +790,20 @@ class StaticCompilationResult(BaseCompilationResult):
789	"""Get standard value"""	790	"""Get standard value"""
790	return self.base_standard_get()	791	return self.base_standard_get()
791		792
792	- def apply(self, testcase: "ttk.TestcaseOp"):	793	+ def apply(self, testcase: "TestcaseOp"):
793	pass	794	pass
794		795
795	def write_json(self, path: Optional[str]):	796	def write_json(self, path: Optional[str]):
796	"""Write compile info json"""	797	"""Write compile info json"""
797	json_parsed = super().get_json()	798	json_parsed = super().get_json()
798	- with open(pathlib.Path(path, "%s.ttk" % self.kernel_name), "w+", encoding="UTF-8") as f:	799	+ with open(pathlib.Path(path, f"{self.kernel_name}.ttk"), "w+", encoding="UTF-8") as f:
799	f.write(json.dumps(json_parsed, indent=4))	800	f.write(json.dumps(json_parsed, indent=4))
800		801
801		802
802	class ConstCompilationResult(StaticCompilationResult):	803	class ConstCompilationResult(StaticCompilationResult):
803	"""For const Compilation"""	804	"""For const Compilation"""
804		805
805	- def apply(self, testcase: "ttk.TestcaseOp"):	806	+ def apply(self, testcase: "TestcaseOp"):
806	"""Apply result to testcase"""	807	"""Apply result to testcase"""
807	if testcase.dyn_func_params is None:	808	if testcase.dyn_func_params is None:
808	testcase.dyn_func_params = self.func_params	809	testcase.dyn_func_params = self.func_params
	@@ -813,7 +814,7 @@ class ConstCompilationResult(StaticCompilationResult):
813	class BinaryCompilationResult(DynamicCompilationResult):	814	class BinaryCompilationResult(DynamicCompilationResult):
814	"""For Binary Compilation"""	815	"""For Binary Compilation"""
815		816
816	- def apply(self, testcase: "ttk.TestcaseOp"):	817	+ def apply(self, testcase: "TestcaseOp"):
817	"""Apply result to testcase"""	818	"""Apply result to testcase"""
818	if testcase.dyn_func_params is None:	819	if testcase.dyn_func_params is None:
819	testcase.dyn_func_params = self.func_params	820	testcase.dyn_func_params = self.func_params

Attk/utilities/table_reader.py+102-0

@@ -0,0 +1,102 @@
	1	+#!/usr/bin/env python3
	2	+# -- coding: UTF-8 --
	3	+# Copyright (c) 2026 Huawei Technologies Co., Ltd.
	4	+# This program is free software, you can redistribute it and/or modify it under the terms and conditions of
	5	+# CANN Open Software License Agreement Version 2.0 (the "License").
	6	+# Please refer to the License for details. You may not use this file except in compliance with the License.
	7	+# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
	8	+# INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
	9	+# See LICENSE in the root of the software repository for the full text of the License.
	10	+"""
	11	+Unified table reader.
	12	+
	13	+Reads CSV and XLSX inputs into a (header, rows) pair of stripped strings,
	14	+so downstream testcase parsing is format-agnostic. CSV path mirrors the
	15	+original UniversalTestcaseFactory._read_csv semantics; XLSX path coerces
	16	+every cell to str (None -> "") to match CSV's text-only behavior.
	17	+"""
	18	+
	19	+__all__ = ["read_table", "read_csv_rows", "read_xlsx_rows", "resolved_sheet"]
	20	+
	21	+import csv
	22	+
	23	+
	24	+def read_csv_rows(fileobj) -> tuple:
	25	+ """Read a CSV file object into (header, rows).
	26	+
	27	+ Each cell is stripped; empty lines are dropped; the first non-empty
	28	+ line is the header. Matches the legacy _read_csv behavior exactly.
	29	+ """
	30	+ rows = []
	31	+ for row in csv.reader(fileobj):
	32	+ row = [column.strip() for column in row]
	33	+ if row:
	34	+ rows.append(row)
	35	+ if not rows:
	36	+ raise ValueError("Empty table: no rows found")
	37	+ return rows[0], rows[1:]
	38	+
	39	+
	40	+def read_xlsx_rows(path, sheet=None) -> tuple:
	41	+ """Read an .xlsx workbook into (header, rows) of stripped strings.
	42	+
	43	+ sheet: worksheet name; default first worksheet. Cells are coerced to
	44	+ str (None -> "") and stripped. Fully-empty rows are dropped, matching
	45	+ the CSV reader's empty-line filtering. data_only=True reads cached
	46	+ formula values; numeric cells become their decimal string form, so
	47	+ shape/dtype columns should be formatted as Text in the workbook.
	48	+ """
	49	+ try:
	50	+ import openpyxl
	51	+ except ImportError as exc: # pragma: no cover
	52	+ raise ImportError("openpyxl is required to read .xlsx inputs; install it with `pip install openpyxl`") from exc
	53	+
	54	+ wb = openpyxl.load_workbook(path, read_only=True, data_only=True)
	55	+ try:
	56	+ if sheet is not None:
	57	+ if sheet not in wb.sheetnames:
	58	+ raise ValueError(f"Sheet '{sheet}' not found in {path}; available: {wb.sheetnames}")
	59	+ ws = wb[sheet]
	60	+ else:
	61	+ ws = wb[wb.sheetnames[0]]
	62	+ rows = []
	63	+ for raw in ws.iter_rows(values_only=True):
	64	+ row = [("" if cell is None else str(cell)).strip() for cell in raw]
	65	+ if any(cell != "" for cell in row):
	66	+ rows.append(row)
	67	+ if not rows:
	68	+ sheet_name = sheet or wb.sheetnames[0]
	69	+ raise ValueError(f"Empty sheet '{sheet_name}' in {path}")
	70	+ return rows[0], rows[1:]
	71	+ finally:
	72	+ wb.close()
	73	+
	74	+
	75	+def read_table(path, sheet=None) -> tuple:
	76	+ """Dispatch by file suffix: .xlsx/.xlsm -> openpyxl, otherwise CSV."""
	77	+ if path.lower().endswith((".xlsx", ".xlsm")):
	78	+ return read_xlsx_rows(path, sheet)
	79	+ with open(path, newline="", encoding="utf-8") as f:
	80	+ return read_csv_rows(f)
	81	+
	82	+
	83	+def resolved_sheet(path, sheet=None):
	84	+ """Return the worksheet name that will actually be read.
	85	+
	86	+ A specified sheet is returned as-is. For xlsx without an explicit
	87	+ sheet, returns the first worksheet name (the default read target).
	88	+ For csv, returns None (no worksheet concept).
	89	+ """
	90	+ if not path.lower().endswith((".xlsx", ".xlsm")):
	91	+ return None
	92	+ if sheet is not None:
	93	+ return sheet
	94	+ try:
	95	+ import openpyxl
	96	+ except ImportError as exc: # pragma: no cover
	97	+ raise ImportError("openpyxl is required to read .xlsx inputs; install it with `pip install openpyxl`") from exc
	98	+ wb = openpyxl.load_workbook(path, read_only=True, data_only=True)
	99	+ try:
	100	+ return wb.sheetnames[0] if wb.sheetnames else None
	101	+ finally:
	102	+ wb.close()

	@@ -56,6 +56,7 @@ def _run(argv, captured):
56		56
57	# -- 子命令注册 --------------------------------------------------------------	57	# -- 子命令注册 --------------------------------------------------------------
58		58
		59	+
59	class TestSubcommandRegistration:	60	class TestSubcommandRegistration:
60	def test_four_modes_registered(self):	61	def test_four_modes_registered(self):
61	"""四个子命令 kernel/geir/aclnn/e2e 均已注册。"""	62	"""四个子命令 kernel/geir/aclnn/e2e 均已注册。"""
	@@ -73,6 +74,7 @@ class TestSubcommandRegistration:
73		74
74	# -- kernel 模式 -------------------------------------------------------------	75	# -- kernel 模式 -------------------------------------------------------------
75		76
		77	+
76	class TestKernelMode:	78	class TestKernelMode:
77	def test_sets_test_mode_op(self, captured):	79	def test_sets_test_mode_op(self, captured):
78	"""kernel 子命令将 test_mode 设为 'op'。"""	80	"""kernel 子命令将 test_mode 设为 'op'。"""
	@@ -101,6 +103,7 @@ class TestKernelMode:
101		103
102	# -- geir 模式 --------------------------------------------------------------	104	# -- geir 模式 --------------------------------------------------------------
103		105
		106	+
104	class TestGeirMode:	107	class TestGeirMode:
105	def test_sets_test_mode_geir(self, captured):	108	def test_sets_test_mode_geir(self, captured):
106	"""geir 子命令将 test_mode 设为 'geir'。"""	109	"""geir 子命令将 test_mode 设为 'geir'。"""
	@@ -126,6 +129,7 @@ class TestGeirMode:
126		129
127	# -- aclnn 模式 -------------------------------------------------------------	130	# -- aclnn 模式 -------------------------------------------------------------
128		131
		132	+
129	class TestAclnnMode:	133	class TestAclnnMode:
130	def test_sets_test_mode_aclnn(self, captured):	134	def test_sets_test_mode_aclnn(self, captured):
131	"""aclnn 子命令将 test_mode 设为 'aclnn'。"""	135	"""aclnn 子命令将 test_mode 设为 'aclnn'。"""
	@@ -135,6 +139,7 @@ class TestAclnnMode:
135		139
136	# -- e2e 模式 ---------------------------------------------------------------	140	# -- e2e 模式 ---------------------------------------------------------------
137		141
		142	+
138	class TestE2eMode:	143	class TestE2eMode:
139	def test_sets_test_mode_framework_api(self, captured):	144	def test_sets_test_mode_framework_api(self, captured):
140	"""e2e 子命令将 test_mode 设为 'framework-api'。"""	145	"""e2e 子命令将 test_mode 设为 'framework-api'。"""
	@@ -159,6 +164,7 @@ class TestE2eMode:
159		164
160	# -- run_with_switches 分派 --------------------------------------------------	165	# -- run_with_switches 分派 --------------------------------------------------
161		166
		167	+
162	class TestRunWithSwitchesDispatch:	168	class TestRunWithSwitchesDispatch:
163	@pytest.mark.parametrize(	169	@pytest.mark.parametrize(
164	"test_mode, instance_path",	170	"test_mode, instance_path",
	@@ -180,7 +186,7 @@ class TestRunWithSwitchesDispatch:
180	monkeypatch.setattr("ttk.core_modules.tbe_logging.default_logging_config", lambda **kw: None)	186	monkeypatch.setattr("ttk.core_modules.tbe_logging.default_logging_config", lambda **kw: None)
181	monkeypatch.setattr("ttk.utilities.set_process_name", lambda: None)	187	monkeypatch.setattr("ttk.utilities.set_process_name", lambda: None)
182	monkeypatch.setattr("ttk.utilities.set_thread_name", lambda: None)	188	monkeypatch.setattr("ttk.utilities.set_thread_name", lambda: None)
183	- monkeypatch.setattr("ttk.cli.bridge._detect_framework_from_csv", lambda files: "torch")	189	+ monkeypatch.setattr("ttk.cli.bridge._detect_framework_from_csv", lambda files, sheet=None: "torch")
184		190
185	with patch(instance_path) as mock_cls:	191	with patch(instance_path) as mock_cls:
186	run_with_switches(sw)	192	run_with_switches(sw)