Closed JMcarrot closed 4 months ago
@JMcarrot Thank you for your question. Can you provide the error log so I can check the details? Basically, EPICKITCHENS.AUDIO_DATA_FILE is an absolute path that pointing the dataset source hdf5 file like "/data_ssd/DATA/EPIC-Kitchens-100-hdf5/EPIC-KITCHENS-100_audio.hdf5". You can refer hdf5 dataset generation guideline via the https://github.com/epic-kitchens/epic-sounds-annotations/tree/main/src. I hope this could help you to solve it.
Thanks for your reply!
My error log is:
`[05/16 06:36:25][INFO] checkpoint.py: 156: Loading network weights from /data/CJM/CVPR/AudioInceptionNeXt-main/output/checkpoints/checkpoint_best.pyth.
[05/16 06:36:25][INFO] epicsound.py: 40: Constructing EPIC-SOUND Audio test...
[05/16 06:36:26][INFO] epicsound.py: 74: Constructing epicsound dataloader (size: 40175) from ['EPIC_Sounds_validation.pkl']
[05/16 06:36:26][INFO] test_net.py: 181: Testing model for 1256 iterations
Process SpawnProcess-2:
Traceback (most recent call last):
File "
-- Process 1 terminated with the following error:
Traceback (most recent call last):
File "/opt/conda/envs/audioIN/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 19, in _wrap
fn(i, args)
File "/data/CJM/CVPR/AudioInceptionNeXt-main/slowfast/utils/multiprocessing.py", line 60, in run
ret = func(cfg)
File "/data/CJM/CVPR/AudioInceptionNeXt-main/tools/test_net.py", line 218, in test
test_meter, preds, preds_clips, labels, metadata = perform_test(test_loader, model, test_meter, cfg, writer)
File "/opt/conda/envs/audioIN/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 26, in decorate_context
return func(args, **kwargs)
File "/data/CJM/CVPR/AudioInceptionNeXt-main/tools/test_net.py", line 50, in perform_test
for cur_iter, (inputs, labels, audio_idx, meta) in enumerate(test_loader):
File "/opt/conda/envs/audioIN/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 435, in next
data = self._next_data()
File "/opt/conda/envs/audioIN/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1085, in _next_data
return self._process_data(data)
File "/opt/conda/envs/audioIN/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1111, in _process_data
data.reraise()
File "/opt/conda/envs/audioIN/lib/python3.8/site-packages/torch/_utils.py", line 428, in reraise
raise self.exc_type(msg)
OSError: Caught OSError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/opt/conda/envs/audioIN/lib/python3.8/site-packages/torch/utils/data/_utils/worker.py", line 198, in _worker_loop
data = fetcher.fetch(index)
File "/opt/conda/envs/audioIN/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/opt/conda/envs/audioIN/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in
and my command is :
python tools/run_net.py --cfg configs/EPIC-SOUND-416x128/AudioInceptionNeXt.yaml --init_method tcp://localhost:9997 \ NUM_GPUS 2 \ OUTPUT_DIR /data/CJM/CVPR/AudioInceptionNeXt-main/test \ EPICKITCHENS.AUDIO_DATA_FILE /data/CJM/CVPR/AudioInceptionNeXt-main/EPIC_audio.hdf5 \ EPICKITCHENS.ANNOTATIONS_DIR /data/CJM/CVPR/AudioInceptionNeXt-main/epic-sounds-annotations-main/ \ TRAIN.ENABLE False \ TEST.ENABLE True \ TEST.CHECKPOINT_FILE_PATH /data/CJM/CVPR/AudioInceptionNeXt-main/output/checkpoints/checkpoint_best.pyth
@JMcarrot, I would like to suggest you check two things first. First, I noticed that there is an error coming out with pkl library. You can try reading the test.pkl file and train.pkl file independently to see if any errors occur. Second, you can read the EPIC_audio.hdf5 independently to see if you can read the dataset properly.
I will check it, thank you!!
Thanks for open-sourcing your code! I have fine-tuning from VGG-Sound pretrained model,but when I run the command to validate the model, the error occurs. The EPICKITCHENS.AUDIO_DATA_FILE parameter is an absolute path and this path can used in train command.