No module named scripts.inference

[ffhelly]

python -m scripts.inference --inference_config configs/inference/test.yaml

error message : No module named scripts.inference

but , i used pip install scripts ， and successed this install.

its same error message。

[vincentWuK]

cd scripts and create empty _ _ init _ _ .py

[vincentWuK]

or just try to run it by python scripts/inference --inference_config configs/inference/test.yaml

[ffhelly]

or just try to run it by python scripts/inference --inference_config configs/inference/test.yaml

thx,bro.

I tried running the program using this method and eventually found that my MMCV framework doesn't seem to match. My current CUDA version is 12.3, but the MMCV prompt version is 11. Is it necessary to downgrade CUDA?

The detected CUDA version (12.3) mismatches the version that was used to compile
PyTorch (11.7). Please make sure to use the same CUDA versions.

[vincentWuK]

or just try to run it by python scripts/inference --inference_config configs/inference/test.yaml

thx,bro.

I tried running the program using this method and eventually found that my MMCV framework doesn't seem to match. My current CUDA version is 12.3, but the MMCV prompt version is 11. Is it necessary to downgrade CUDA?

The detected CUDA version (12.3) mismatches the version that was used to compile PyTorch (11.7). Please make sure to use the same CUDA versions.

you can first try to use conda for managing cuda environment. if not works, try to use devcontainer in vscode.

[vincentWuK]

conda create -> conda activate -> python3 -m pip install pytorch and pytorch-cuda, ....

[ffhelly]

conda create -> conda activate -> python3 -m pip install pytorch and pytorch-cuda, ....

Do you mean to create a new conda environment, point to the corresponding pytorch version (11), and then deploy it?
Will the 12.2 version I saw in Nvidia smi conflict with the current new conda version environment? bro