Skip to content

Testing LoViT using Cholec80 data

Benckendorff, D.M. von (Dimitri, Student M-CS) requested to merge Dimitri into main

Tried to retrieve spatial features from video input

IMPLEMENTED:

  • version specification for timm and torch
  • video preprocessing to get tensor in shape (Batch, Channels, Frame, Height, Width)
  • spatial feature extraction to get tensor in shape (Batch, Classes, Frame, 1, 1)
  • spatial feature extraction for the first 10 frames of video01.mp4

SOLVED AttributeError: 'GELU' object has no attribute 'approximate' fixed by replacing gelu layer and removing approximation in gelu layers.

UNRESOLVED! OutOfMemoryError: CUDA out of memory. Tried to allocate 24.30 GiB (GPU 0; 14.65 GiB total capacity; 327.30 MiB already allocated; 3.21 GiB free; 382.00 MiB 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 has not been fixed yet.

Merge request reports