0
代码
代码
Issues
Pull Requests
讨论
Wiki
分析
0
ascend-robotascend-robotcache_compile模式下TensorifyScalarRestartAnalysis导致parameters重复保存
9e85f623创建于 3 小时前历史提交
文件最后提交记录最后更新时间
CMakeLists.txt
ge自动converter异常日志优化 Co-authored-by: luanchaowei<luanchaowei3@huawei.com> # message auto-generated for no-merge-commit merge: !2816 merge master into master ge自动converter异常日志优化 Created-by: luanchaowei Commit-by: luanchaowei Merged-by: ascend-robot Description: ge自动converter异常日志优化 See merge request: Ascend/torchair!28161 个月前
allowlist_for_publicAPI.json
add npu_format_cast_via_cpu + add quantmatmul lite converter # message auto-generated for no-merge-commit merge: !3092 merge master into master add npu_format_cast_via_cpu + add quantmatmul lite converter Created-by: zihan0007 Commit-by: zihan0007 Merged-by: ascend-robot Description: 合入来源： torchair 新增npu_format_cast_via_cpu接口： 新增npu_format_cast_via_cpu接口 支持npuTensor在cpu将数据排布从Nd转为Nz，当前支持int8/uint8 Nd输入转为Fractal_nz输出 ge_pass增加QuantBatchMatmulV3前的Bitcast和Transpose算子互换位置的能力，方便后续Transpose融合 增加lite converter支持QuantBatchMatmulV3 A8W4（int4）场景的converter逻辑 See merge request: Ascend/torchair!30928 天前
torch_npu_schema.json
add npu_format_cast_via_cpu + add quantmatmul lite converter # message auto-generated for no-merge-commit merge: !3092 merge master into master add npu_format_cast_via_cpu + add quantmatmul lite converter Created-by: zihan0007 Commit-by: zihan0007 Merged-by: ascend-robot Description: 合入来源： torchair 新增npu_format_cast_via_cpu接口： 新增npu_format_cast_via_cpu接口 支持npuTensor在cpu将数据排布从Nd转为Nz，当前支持int8/uint8 Nd输入转为Fractal_nz输出 ge_pass增加QuantBatchMatmulV3前的Bitcast和Transpose算子互换位置的能力，方便后续Transpose融合 增加lite converter支持QuantBatchMatmulV3 A8W4（int4）场景的converter逻辑 See merge request: Ascend/torchair!30928 天前
torchair_aclgraph_reinplace_st.py
【bugfix】 input_inplace_pass在对输入self.copy_时会报错 Co-authored-by: wangkai<wangkai579@huawei.com> # message auto-generated for no-merge-commit merge: !2930 merge master into master 【bugfix】 input_inplace_pass在对输入self.copy_时会报错 Created-by: mihudan Commit-by: wangkai Merged-by: ascend-robot Description: 【bugfix】 input_inplace_pass在对输入self.copy_时会报错 ``` _reinplace_input_mutated_ops(self.fx_graph) python3.10/site-packages/torch_npu/dynamo/npugraph_ex/_utils/graph_utils.py", line 221, in wrapper result = func(*args, **kwargs) python3.10/site-packages/torch_npu/dynamo/npugraph_ex/_acl_concrete_graph/graph_pass.py", line 571, in _reinplace_input_mutated_ops _mutated_input_reinplace(gm) python3.10/site-packages/torch_npu/dynamo/npugraph_ex/_acl_concrete_graph/graph_pass.py", line 462, in _mutated_input_reinplace mutated_arg = node.args[0] IndexError: tuple index out of range ``` 报错信息 ``` class Network(nn.Module): def forward(self, x): z = x.copy_(x) return z ``` 用户脚本 ``` print(graph) graph(): %arg0_1 : [num_users=1] = placeholder[target=arg0_1] %copy_ : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%arg0_1, %arg0_1), kwargs = {}) return (copy_,) ``` fx图 See merge request: Ascend/torchair!29301 个月前
torchair_aclgraph_st.py
fix: wait_tensor is not implemented error message Co-authored-by: xuzhenhua2025<xuzhenhua4@huawei.com> # message auto-generated for no-merge-commit merge: !3052 merge master into master fix: wait_tensor is not implemented error message Created-by: xuzhenhua2025 Commit-by: xuzhenhua2025 Merged-by: ascend-robot Description: fix wait_tensor is not implemented error message See merge request: Ascend/torchair!305220 天前
torchair_auto_converter_functional_st.py
pre-commit by rule=trailing-whitespace Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2931 merge clean_code into master pre-commit by rule=trailing-whitespace Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: pre-commit-clean-code-rule=trailing-whitespace See merge request: Ascend/torchair!29311 个月前
torchair_cache_compile_st.py
update code format Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2944 merge update_log into master update code format Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: update code format See merge request: Ascend/torchair!29441 个月前
torchair_converter_error_message_st.py
pre-commit by rule=trailing-whitespace Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2931 merge clean_code into master pre-commit by rule=trailing-whitespace Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: pre-commit-clean-code-rule=trailing-whitespace See merge request: Ascend/torchair!29311 个月前
torchair_customop_st.py
update code format Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2944 merge update_log into master update code format Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: update code format See merge request: Ascend/torchair!29441 个月前
torchair_debug_log_st.py
add npu_format_cast_via_cpu + add quantmatmul lite converter # message auto-generated for no-merge-commit merge: !3092 merge master into master add npu_format_cast_via_cpu + add quantmatmul lite converter Created-by: zihan0007 Commit-by: zihan0007 Merged-by: ascend-robot Description: 合入来源： torchair 新增npu_format_cast_via_cpu接口： 新增npu_format_cast_via_cpu接口 支持npuTensor在cpu将数据排布从Nd转为Nz，当前支持int8/uint8 Nd输入转为Fractal_nz输出 ge_pass增加QuantBatchMatmulV3前的Bitcast和Transpose算子互换位置的能力，方便后续Transpose融合 增加lite converter支持QuantBatchMatmulV3 A8W4（int4）场景的converter逻辑 See merge request: Ascend/torchair!30928 天前
torchair_eager_st.py
add_rms_norm支持inplace pass Co-authored-by: JieTang<tangjie66@huawei.com> # message auto-generated for no-merge-commit merge: !2461 merge master-16 into master add_rms_norm支持inplace pass Created-by: tangjie66 Commit-by: JieTang Merged-by: ascend-robot Description: add_rms_norm支持inplace pass 【新功能提交】 背景 torchair中有针对模型中间节点，将节点中包含的非原地算子替换未原地算子的pass，目前只支持单原地算子的reinplace，例如从add(a, b)变为a.add_(b) 当存在中间节点自定义的多原地算子时，不会处理 变更内容 本次PR完成以下内容： 新增对 中间节点自定义的多原地算子 的处理 新增自定义多原地算子add_rms_norm_v2 新增相关测试用例 See merge request: Ascend/torchair!24614 个月前
torchair_explicit_order_st.py
GeOption支持确定性级别 Co-authored-by: zhukkk<zhuke11@huawei.com> # message auto-generated for no-merge-commit merge: !2586 merge master_0202 into master GeOption支持确定性级别 Created-by: zhukkk Commit-by: zhukkk Merged-by: ascend-robot Description: GeOption支持确定性级别 See merge request: Ascend/torchair!25863 个月前
torchair_export_st.py
[fix]修改lite converter覆盖注册方式 Co-authored-by: geyi<geyi2@huawei.com> # message auto-generated for no-merge-commit merge: !2907 merge master_fix_lite into master [fix]修改lite converter覆盖注册方式 Created-by: gleaming-spark Commit-by: geyi Merged-by: ascend-robot Description: [fix]修改lite converter覆盖注册方式 See merge request: Ascend/torchair!29071 个月前
torchair_fx_pattern_st.py
rollback pattern pass for ge Co-authored-by: yurongkun<yurongkun@huawei.com> # message auto-generated for no-merge-commit merge: !2526 merge patternpassupdate into master rollback pattern pass for ge Created-by: yurongkun Commit-by: yurongkun Merged-by: ascend-robot Description: rollback pattern pass for ge See merge request: Ascend/torchair!25264 个月前
torchair_gear_st.py
!1976 [bugfix]fix 2.6 st error Merge pull request !1976 from 陈威亨/master 8 个月前
torchair_hcom_st.py
move patch for traceable_collective_remaps to init Co-authored-by: huangfei<huangfei32@huawei.com> # message auto-generated for no-merge-commit merge: !2598 merge master_02 into master move patch for traceable_collective_remaps to init Created-by: dddxhh Commit-by: huangfei Merged-by: ascend-robot Description: move patch for traceable_collective_remaps to init See merge request: Ascend/torchair!25983 个月前
torchair_loglevel_st.py
pre-commit by rule=trailing-whitespace Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2931 merge clean_code into master pre-commit by rule=trailing-whitespace Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: pre-commit-clean-code-rule=trailing-whitespace See merge request: Ascend/torchair!29311 个月前
torchair_npugraph_ex_cache_st.py
cache_compile模式下TensorifyScalarRestartAnalysis导致parameters重复保存 Co-authored-by: weizhaoyang<weizhaoyang1@huawei.com> # message auto-generated for no-merge-commit merge: !3150 merge wzy0603 into master cache_compile模式下TensorifyScalarRestartAnalysis导致parameters重复保存 Created-by: weizhaoyang Commit-by: weizhaoyang Merged-by: ascend-robot Description: cache_compile模式下TensorifyScalarRestartAnalysis导致parameters重复保存 See merge request: Ascend/torchair!31503 小时前
torchair_npugraph_ex_compile_st.py
force eager add clone input and output Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !3098 merge update_eager into master force eager add clone input and output Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: force eager add clone input and output See merge request: Ascend/torchair!30988 小时前
torchair_print_detail_st.py
!2085 check tensor_detail argument of npu_print() type Merge pull request !2085 from wangkai/master 7 个月前
torchair_print_st.py
GeOption支持确定性级别 Co-authored-by: zhukkk<zhuke11@huawei.com> # message auto-generated for no-merge-commit merge: !2586 merge master_0202 into master GeOption支持确定性级别 Created-by: zhukkk Commit-by: zhukkk Merged-by: ascend-robot Description: GeOption支持确定性级别 See merge request: Ascend/torchair!25863 个月前
torchair_public_bindings_st.py
Support independent packaging and deployment of npugraph_ex Co-authored-by: xuzhenhua2025<xuzhenhua4@huawei.com> # message auto-generated for no-merge-commit merge: !2689 merge master into master Support independent packaging and deployment of npugraph_ex Created-by: xuzhenhua2025 Commit-by: xuzhenhua2025 Merged-by: ascend-robot Description: 1. 通信算子注册前增加判断，避免重复注册 2. 在CI torch_npu_schema.json添加npugraph_ex函数签名 See merge request: Ascend/torchair!26892 个月前
torchair_scope_st.py
[feat]npu_stream_switch接口新增enable_inner_parallel参数 Co-authored-by: geyi<geyi2@huawei.com> # message auto-generated for no-merge-commit merge: !2956 merge mater_feat_parallel into master [feat]npu_stream_switch接口新增enable_inner_parallel参数 Created-by: gleaming-spark Commit-by: geyi Merged-by: ascend-robot Description: [feat]npu_stream_switch接口新增enable_inner_parallel参数 See merge request: Ascend/torchair!29561 个月前
torchair_st.py
add npu_format_cast_via_cpu + add quantmatmul lite converter # message auto-generated for no-merge-commit merge: !3092 merge master into master add npu_format_cast_via_cpu + add quantmatmul lite converter Created-by: zihan0007 Commit-by: zihan0007 Merged-by: ascend-robot Description: 合入来源： torchair 新增npu_format_cast_via_cpu接口： 新增npu_format_cast_via_cpu接口 支持npuTensor在cpu将数据排布从Nd转为Nz，当前支持int8/uint8 Nd输入转为Fractal_nz输出 ge_pass增加QuantBatchMatmulV3前的Bitcast和Transpose算子互换位置的能力，方便后续Transpose融合 增加lite converter支持QuantBatchMatmulV3 A8W4（int4）场景的converter逻辑 See merge request: Ascend/torchair!30928 天前
torchair_st_stub_aclgraph_utils.py
st用例修改，以使2.6版本torch能区分不同场景的pattern Co-authored-by: JieTang<tangjie66@huawei.com> # message auto-generated for no-merge-commit merge: !2785 merge testcase into master st用例修改，以使2.6版本torch能区分不同场景的pattern Created-by: tangjie66 Commit-by: JieTang Merged-by: ascend-robot Description: 2.6版本pattern校验重复会根据pattern是否相等进行区分， 而当前st用例里对相应算子的patch直接返回了tensor， 会导致部分不同场景生成的pattern一样，报pattern重复 See merge request: Ascend/torchair!27852 个月前
torchair_st_utils.py
pre-commit by rule=trailing-whitespace Co-authored-by: guanlongfeng<guanlongfeng1@huawei.com> # message auto-generated for no-merge-commit merge: !2931 merge clean_code into master pre-commit by rule=trailing-whitespace Created-by: glf2016 Commit-by: guanlongfeng Merged-by: ascend-robot Description: pre-commit-clean-code-rule=trailing-whitespace See merge request: Ascend/torchair!29311 个月前