Update skip layer norm #22719
+218
−83
Merged
Update skip layer norm #22719
Azure Pipelines / Linux GPU CI Pipeline
succeeded
Nov 12, 2024 in 2h 35m 40s
Build #20241111.28 had test failures
Details
- Failed: 1 (0.00%)
- Passed: 47,908 (98.78%)
- Other: 592 (1.22%)
- Total: 48,501
Annotations
Check failure on line 1 in cuda_self_attn_fp16
azure-pipelines / Linux GPU CI Pipeline
cuda_self_attn_fp16
/onnxruntime_src/onnxruntime/test/providers/checkers.cc:437
The difference between f_expected[i] and f_actual[i] is 0.02154541015625, which exceeds tolerance, where
f_expected[i] evaluates to 0.03668212890625,
f_actual[i] evaluates to 0.0582275390625, and
tolerance evaluates to 0.020036682486534119.
i:2357
Google Test trace:
/onnxruntime_src/onnxruntime/test/providers/checkers.cc:568: provider type: CUDAExecutionProvider
/onnxruntime_src/onnxruntime/test/providers/base_tester.cc:841: registered execution providers: CUDAExecutionProvider
/onnxruntime_src/onnxruntime/test/common/random_generator.h:50: ORT test random seed: 4210570007
/onnxruntime_src/onnxruntime/test/common/tensor_op_test_utils.cc:12: ORT test random seed with fixed pattern tensor generator
Raw output
/onnxruntime_src/onnxruntime/test/providers/checkers.cc:437
The difference between f_expected[i] and f_actual[i] is 0.02154541015625, which exceeds tolerance, where
f_expected[i] evaluates to 0.03668212890625,
f_actual[i] evaluates to 0.0582275390625, and
tolerance evaluates to 0.020036682486534119.
i:2357
Google Test trace:
/onnxruntime_src/onnxruntime/test/providers/checkers.cc:568: provider type: CUDAExecutionProvider
/onnxruntime_src/onnxruntime/test/providers/base_tester.cc:841: registered execution providers: CUDAExecutionProvider
/onnxruntime_src/onnxruntime/test/common/random_generator.h:50: ORT test random seed: 4210570007
/onnxruntime_src/onnxruntime/test/common/tensor_op_test_utils.cc:12: ORT test random seed with fixed pattern tensor generator
Loading