How to supress using GPU card while using OnnxCreate()

New comment
 

Hi Fellow

I am facing this error and any help to avoid this will be highly appreciable. Seems hardware compatibility issue:

___________________________________________________________________________________________________________________________________________________________________________

mONNXHandleLong = OnnxCreate(mONNXFileLong, ONNX_DEFAULT | ONNX_COMMON_FOLDER);

mONNXHandleShort = OnnxCreate(mONNXFileShort, ONNX_DEFAULT | ONNX_COMMON_FOLDER);

___________________________________________________________________________________________________________________________________________________________________________

BTCUSD[H4] [CiSuperTrend::getSTBuffer] ERROR[4806] getting iSuperTrend value at Index[2]

BTCUSD[H1] [CS10101ST::updateFeaturesInDB] COMMITTED 1 bars successfully | Errors[0]

ONNX: 1 Memcpy nodes are added to the graph e9a776d8e86f4574a6c1ab309a60a86e for CUDAExecutionProvider. It might have negative impact on performance (including unable to run CUDA graph). Set session_options.log_severity_level=1 to see the detail logs before this message.

ONNX: Some nodes were not assigned to the preferred execution providers which may or may not have an negative impact on performance. e.g. ORT explicitly assigns shape related ops to CPU to improve perf.

ONNX: Rerunning with verbose output on a non-minimal build will show node assignments.

ONNX: Exception during initialization: E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_call.cc:129 onnxruntime::CudaCall E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_call.cc:121 onnxruntime::CudaCall CUBLAS failure 8: CUBLAS_STATUS_ARCH_MISMATCH ; GPU=0 ; hostname=HPZ230 ; file=E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_execution_provider.cc ; line=228 ; expr=cublasCreate(&cublas_handle

ONNX: create session failed with 1 status 'Exception during initialization: E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_call.cc:129 onnxruntime::CudaCall E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_call.cc:121 onnxruntime::CudaCall CUBLAS failure 8: CUBLAS_STATUS_ARCH_MISMATCH ; GPU=0 ; hostname=HPZ230 ; file=E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_execution_provider.cc ; line=2

BTCUSD[H1] [CS10101ST::loadONNXModels] ERROR[5800] Failed to load Long ONNX: ONNXModel\CS10101ST\CS10101ST_BTCUSD_H1_Long.onnx

BTCUSD[H1] [CS10101ST::CS10101ST] WARNING: ONNX models failed to load — EA will not trade

Regards and Thanks in advance


 
Anil Varma:

I am facing this error and any help to avoid this will be highly appreciable. Seems hardware compatibility issue...

"The model runs fine with CUDA Graphs after upgrading to ORT 1.16.1..."

[Performance] How can I forcefully assign nodes to CUDA EP? · Issue #17930 · microsoft/onnxruntime
[Performance] How can I forcefully assign nodes to CUDA EP? · Issue #17930 · microsoft/onnxruntime
  • microsoft
  • github.com
Describe the issue Similar to #16863, I have a model which are assigned to CPU with the following warning: [W:onnxruntime:, session_state.cc:1030 VerifyEachNodeIsAssignedToAnEp] Some nodes were not...
 
Anil Varma:

Hi Fellow

I am facing this error and any help to avoid this will be highly appreciable. Seems hardware compatibility issue:

___________________________________________________________________________________________________________________________________________________________________________

mONNXHandleLong = OnnxCreate(mONNXFileLong, ONNX_DEFAULT | ONNX_COMMON_FOLDER);

mONNXHandleShort = OnnxCreate(mONNXFileShort, ONNX_DEFAULT | ONNX_COMMON_FOLDER);

___________________________________________________________________________________________________________________________________________________________________________

BTCUSD[H4] [CiSuperTrend::getSTBuffer] ERROR[4806] getting iSuperTrend value at Index[2]

BTCUSD[H1] [CS10101ST::updateFeaturesInDB] COMMITTED 1 bars successfully | Errors[0]

ONNX: 1 Memcpy nodes are added to the graph e9a776d8e86f4574a6c1ab309a60a86e for CUDAExecutionProvider. It might have negative impact on performance (including unable to run CUDA graph). Set session_options.log_severity_level=1 to see the detail logs before this message.

ONNX: Some nodes were not assigned to the preferred execution providers which may or may not have an negative impact on performance. e.g. ORT explicitly assigns shape related ops to CPU to improve perf.

ONNX: Rerunning with verbose output on a non-minimal build will show node assignments.

ONNX: Exception during initialization: E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_call.cc:129 onnxruntime::CudaCall E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_call.cc:121 onnxruntime::CudaCall CUBLAS failure 8: CUBLAS_STATUS_ARCH_MISMATCH ; GPU=0 ; hostname=HPZ230 ; file=E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_execution_provider.cc ; line=228 ; expr=cublasCreate(&cublas_handle

ONNX: create session failed with 1 status 'Exception during initialization: E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_call.cc:129 onnxruntime::CudaCall E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_call.cc:121 onnxruntime::CudaCall CUBLAS failure 8: CUBLAS_STATUS_ARCH_MISMATCH ; GPU=0 ; hostname=HPZ230 ; file=E:\workspace\external\onnx\onnx-avx2\onnxruntime\onnxruntime\core\providers\cuda\cuda_execution_provider.cc ; line=2

BTCUSD[H1] [CS10101ST::loadONNXModels] ERROR[5800] Failed to load Long ONNX: ONNXModel\CS10101ST\CS10101ST_BTCUSD_H1_Long.onnx

BTCUSD[H1] [CS10101ST::CS10101ST] WARNING: ONNX models failed to load — EA will not trade

Regards and Thanks in advance

Can you share code and model to check ? In private if you prefer.
New comment