search

pytorch / hub

Submission to https://pytorch.org/hub/
1.39k stars 242 forks source link

Failed in running Tacotron2(NVIDA) #321

Open fengbaoyihao3000 opened 1 year ago

fengbaoyihao3000 commented 1 year ago

I followed the guide to test Demo, but it didn't work and came with 'RuntimeError: CUDA error: CUBLAS_STATUS_NOT_INITIALIZED when calling `cublasCreate(handle)'. os:Windows10 torch:1.10.1 CUDA:11.3

I will appreciate if you can solve this problem

babbangida commented 1 year ago

If u haven’t solved this issue yet Check CUDA Installation: Make sure that CUDA is installed correctly on your system and the versions are compatible. Ensure that you have CUDA 11.3 installed, which matches your GPU's capabilities. Verify GPU Compatibility: Confirm that your GPU is compatible with CUDA 11.3. You can check the official NVIDIA website to ensure your GPU model supports this version. Update GPU Drivers: Update your GPU drivers to the latest version compatible with CUDA 11.3. Visit the NVIDIA website and download the appropriate drivers for your GPU model. Check PyTorch Compatibility: Verify that the version of PyTorch you installed (1.10.1) is compatible with CUDA 11.3. You can refer to the PyTorch documentation or release notes to check for any known compatibility issues between the PyTorch version and CUDA version.

Unicorncosmos commented 1 year ago

I followed the guide to test Demo, but it didn't work and came with 'RuntimeError: CUDA error: CUBLAS_STATUS_NOT_INITIALIZED when calling `cublasCreate(handle)'. os:Windows10 torch:1.10.1 CUDA:11.3

I will appreciate if you can solve this problem

Verify CUDA Installation: Make sure CUDA is properly installed on your system and the PATH environment variable is correctly set. Check that you have the compatible GPU driver installed for CUDA 11.3. Verify PyTorch Compatibility: Ensure that you have installed a version of PyTorch that is compatible with CUDA 11.3. Visit the PyTorch website or the official PyTorch GitHub repository to check the compatible versions. Check GPU Availability: Verify that your GPU is recognized by CUDA and accessible to PyTorch. You can use the torch.cuda.is_available() function to check if PyTorch can detect your GPU. Update GPU Drivers: If your GPU driver is outdated, try updating it to the latest version compatible with CUDA 11.3. Reinstall PyTorch: Uninstall the existing PyTorch installation and reinstall it using a compatible version for CUDA 11.3. Check System Compatibility: Ensure that your Windows 10 version is compatible with CUDA 11.3 and PyTorch. Check for Multiple CUDA Installations: If you have multiple versions of CUDA installed on your system, it can cause conflicts. Make sure you are using the correct version for your setup. Try CPU-only Mode: If running on GPU is not essential for your task, you can try running the code in CPU-only mode by setting the device to 'cpu' instead of 'cuda'