黄巧波

Results 11 comments of 黄巧波

Hi @kfern Thank you for your answer I found that export cmyk image is not supported by exportMethods of Image class Because of that, I have no way to export...

after run `pip install interence` get new error: `ImportError: cannot import name 'YOLOWorld' from 'inference.models'` edit YOLO_WORLD_EfficientSAM.py can resolve the error ```python - from inference.models import YOLOWorld + from inference.models.yolo_world...

> The issue is a memory leak issue in the pytorch build, it's resolved by using the pytorch nightly preview, but that can cause other issues. I did find that...

> FYI, that feature doesn't release cached data. You have to execute empty workflow before free_memory if you want to wipe all. I then ran this workflow and the memory...

> > > FYI, эта функция не освобождает кэшированные данные. Вам нужно выполнить пустой рабочий процесс перед free_memory, если вы хотите стереть все. > > > > > > >...

i find the memory leak is happen in torch function `module._load_from_state_dict`

> > i find the memory leak is happen in torch function `module._load_from_state_dict` > > Do you use torch 2.4.0 on windows? i have tried these versions 2.3.0 on mac....

> ![image](https://private-user-images.githubusercontent.com/77841885/367541930-9c4a04c2-7ee1-47c9-b69d-6310c1f49d4b.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.rIQaywVNuzubjN7YHUVgMNZkMA7JAgkXVdfZejMog2s) > > Seems like the model disposage is the issue here: (This does clear the VRAM but duplicates the cache in ram) > > ``` > model_management.unload_all_models() >...

I guess this unreleased memory is pinned memory, it's managed by [CUDACachingHostAllocator](https://github.com/pytorch/pytorch/blob/ebeab262d90bd5e31a760da0ac7a3399bfb44f61/aten/src/ATen/cuda/CachingHostAllocator.cpp#L236). If we want to clean up this memory, we need to call the `CachingHostAllocator_emptyCache` method. This method can...

sad. `torch._C._host_emptyCache()` method is not work in this case. Because `CheckpointLoaderSimple` may not using pin_memory. However, some third-party nodes use `pin_memory=True` when using `torch.utils.data.DataLoader`. If you use such nodes, you...