search

facebookresearch / SlowFast

PySlowFast: video understanding codebase from FAIR for reproducing state-of-the-art video models.
Apache License 2.0
6.52k stars 1.2k forks source link

Errors When Train Custom Dataset with c2/SLOWFAST_8x8_R101_101_101 #237

Open npduyduc opened 4 years ago

npduyduc commented 4 years ago

I'm now trying to train SLOWFAST_8x8_R101_101_101 with my dataset and meetting error:

RuntimeError: DataLoader worker (pid(s) 2133, 2135, 2138, 2139) exited unexpectedly

But, I can train with C2D_8x8_R50.

And the second question, you can tell me the directory output model when training with google colab? When I continue to train it can load from the last checkpoint but I can find this file with the search tool, and "checkpoints" folder is empty. I have used google colab, run this command to fix errors have space in [My Drive]

!ln -s "/content/drive/My Drive/action_recognition/slowfast/" "/content/"

and run your code

!python "/content/drive/My Drive/action_recognition/slowfast/tools/run_net.py" \
  --cfg "/content/drive/My Drive/action_recognition/slowfast/configs/Kinetics/c2/SLOWFAST_16x8_R101_50_50.yaml" \
  DATA.PATH_TO_DATA_DIR "/content/drive/My Drive/action_recognition/slowfast/dataset" \
  NUM_GPUS 1 \
  TRAIN.BATCH_SIZE 16 \

Can you help me with these issues? Thank you!

bqhuyy commented 4 years ago

I'm now trying to train SLOWFAST_8x8_R101_101_101 with my dataset and meetting error:

RuntimeError: DataLoader worker (pid(s) 2133, 2135, 2138, 2139) exited unexpectedly

But, I can train with C2D_8x8_R50.

And the second question, you can tell me the directory output model when training with google colab? When I continue to train it can load from the last checkpoint but I can find this file with the search tool, and "checkpoints" folder is empty. I have used google colab, run this command to fix errors have space in [My Drive]

!ln -s "/content/drive/My Drive/action_recognition/slowfast/" "/content/"

and run your code

!python "/content/drive/My Drive/action_recognition/slowfast/tools/run_net.py" \
  --cfg "/content/drive/My Drive/action_recognition/slowfast/configs/Kinetics/c2/SLOWFAST_16x8_R101_50_50.yaml" \
  DATA.PATH_TO_DATA_DIR "/content/drive/My Drive/action_recognition/slowfast/dataset" \
  NUM_GPUS 1 \
  TRAIN.BATCH_SIZE 16 \

Can you help me with these issues? Thank you!

I think you should decrease NUM_WORKERS