How to make it faster? model loading and processing can be extremely slow

[alanmun]

Some (not all) workflows are MUCH slower to run, and all models load MUCH slower.

For example, I am running a workflow where I just simply want to upscale, the model is 144MB in size but load times are really bad, like 15 minutes.

I have Win11 Home edition, cu128 image, WSL2 integration, so I've set .wslconfig to have 64GB of system ram and 24 logical threads. I'm using the normal speed ups that I already use. I use a docker volume instead of a windows mount. I have a 4090 and gave it all 24GB to work with.

By comparison, the same comfyui setup on the same windows machine is far faster directly on windows.

[Zalozheng]

bitlocker must close，your host I/O slow. Windows Defender must close.

or make the cache directory persistent

-v $(pwd)/storage-models/hf-hub:/root/.cache/huggingface/hub
-v $(pwd)/storage-models/torch-hub:/root/.cache/torch/hub

[alanmun]

Thanks, I will try the cache directory idea. I am not willing to disable windows defender or bitlocker for security reasons nor do I think you should suggest that people try that...

[alanmun]

It looks like .cache directory was already persisting for me, too