Reputation: 387
I converted a tensorflow saved model to ONNX format using tf2onnx :
python3 -m tf2onnx.convert --saved-model saved_model/ --output onnx/model.onnx --opset 11
The conversion worked fine and I can run inference with the ONNX model using CPU.
I installed onnxruntime-gpu
to run inference with GPU and encountered an error :
RuntimeException: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running Relu node. Name:'FirstStageFeatureExtractor/resnet_v1_101/resnet_v1_101/conv1/Relu' Status Message: /onnxruntime_src/onnxruntime/core/providers/cuda/cuda_call.cc:97 bool onnxruntime::CudaCall(ERRTYPE, const char*, const char*, ERRTYPE, const char*) [with ERRTYPE = cudaError; bool THRW = true] /onnxruntime_src/onnxruntime/core/providers/cuda/cuda_call.cc:91 bool onnxruntime::CudaCall(ERRTYPE, const char*, const char*, ERRTYPE, const char*) [with ERRTYPE = cudaError; bool THRW = true] CUDA failure 2: out of memory ; GPU=0 ; hostname=coincoin; expr=cudaMalloc((void**)&p, size);
Stacktrace:
Stacktrace:
I am the only one using the GPU which is a Titan RTX (24GB of RAM). The model runs fine on GPU using its tensorflow saved model version, with 10GB of the GPU's RAM.
Versions are :
Upvotes: 0
Views: 4195
Reputation: 101
according to your information and Versions, maybe two solutions to solve the problem:
config = tf.ConfigProto()
config.gpu_options.visible_device_list = "0"
config.gpu_options.per_process_gpu_memory_fraction = 0.1
set_session(tf.Session(config=config))
Upvotes: 0