run on Colab with A100 GPU and 40G GPU RAM still got Error

[bingjiw]

run webUI.py on Colab with A100 GPU and 40G GPU RAM still got Error:

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 12.25 GiB (GPU 0; 39.56 GiB total capacity; 21.83 GiB already allocated; 10.45 GiB free; 22.58 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[williamyang1991]

Can you run the example videos on 40GB on Colab?
My example video requires less than 24 GB on my 4090 GPU.

[bingjiw]

Yes, the example of "run_fresco.ipynb" is sucessfully running on Colab with A100 with 40GB GPU RAM.

[williamyang1991]

What is the resolution of the video that causes your OOM?

[bingjiw]

1280x720.mp4 and 640x360.mov

[williamyang1991]

Then it will be resized to 512x896
Compared to 512*512 (example video), it has 1.75X pixels
Then, in our feature optimization, we need to compute the gram matrix, which will be 1.75x1.75=3.0625X memory.
Causing OOM.

Maybe you can use smaller batch size or resize video to smaller resolution.

FRESCO/config/config_carturn.yaml

Line 20 in b3148c2

batch_size: 8

FRESCO/run_fresco.py

Line 170 in 9fe1be7

img = resize_image(frame, 512)

[bingjiw]

thank you