R3gm
Cuda memory error
run in docker
docker run -it -p 7860:7860 --platform=linux/amd64 --gpus all
-e OPENAI_API_KEY="---"
-e YOUR_HF_TOKEN=""
-e ZERO_GPU="TRUE"
-e IS_DEMO="TRUE"
registry.hf.space/r3gm-sonitranslate-translate-audio-of-a-video-content:latest python app_rvc.py
Win10 Cuda 12.8 13th Gen Intel(R) Core(TM) i7-13800H 2.50 GHz 64gb RAM NVIDIA RTX A2000 8GB
Video Fille = 148mb if I upload a file of 10-30 MB then there are no errors
2025-03-12 11:25:55.451209551 [E:onnxruntime:Default, cuda_call.cc:118 CudaCall] CUDA failure 700: an illegal memory access was encountered ; GPU=0 ; hostname=5a259a96ae8d ; file=/onnxruntime_src/onnxruntime/core/providers/cuda/tensor/nonzero_op.cc ; line=77 ; expr=cudaMemcpyAsync( &nonzero_elements, prefix_counts + number_of_blocks - 1, sizeof(int), cudaMemcpyDeviceToHost, Stream(context)); 2025-03-12 11:25:55.451254514 [E:onnxruntime:, sequential_executor.cc:516 ExecuteKernel] Non-zero status code returned while running NonZero node. Name:'/dp/flows.5/NonZero_5' Status Message: CUDA failure 700: an illegal memory access was encountered ; GPU=0 ; hostname=5a259a96ae8d ; file=/onnxruntime_src/onnxruntime/core/providers/cuda/tensor/nonzero_op.cc ; line=77 ; expr=cudaMemcpyAsync( &nonzero_elements, prefix_counts + number_of_blocks - 1, sizeof(int), cudaMemcpyDeviceToHost, Stream(context));
