Luca Wiehe comments

Results 4 comments of


                                            Luca Wiehe

cuda out of memory

In the `v2`-version of the project, I have the same problem despite having 24GB of GPU memory. I trained SAM masks using the `--downsample 4` flag. Since you have the...

RuntimeError: CUDA out of memory. Tried to allocate 11.04 GiB (GPU 0; 47.53 GiB total capacity; 30.83 GiB already allocated; 8.16 GiB free; 34.96 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.

Hey, I have faced the same issue and this is a drawback of this approach. For all pixels in the image, you generate additional features that are then gonna be...

RuntimeError: CUDA out of memory. Tried to allocate 11.04 GiB (GPU 0; 47.53 GiB total capacity; 30.83 GiB already allocated; 8.16 GiB free; 34.96 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.

Before `python train_contrastive_feature.py`, you extract image data using `python extract_segment_everything_masks.py` and `python get_scale.py`. You specify image locations using `--image_root `. If you have a lot of images with minor differences...

ImportError: ImportError: cannot import name 'IGNORE_GIT_FOLDER_PATTERNS' from 'huggingface_hub.utils' ....

This does not work for me. There seems to be a conflict with the transformer package: > ERROR: pip's dependency resolver does not currently take into account all the packages...