Integrate pipeline with activeloop hub #1
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Work performed
After exporting a training dataset in YOLO format from Labelflow and send it as a Hub dataset using labelflow/labelflow#659, I created here a training script to use any Hub dataset for training, instead of a local Coco dataset as done originally. Additionally, I created a benchmark script "speed_test_hub.py" that compares the speed of data loading whether it's local or remote w/ hub. Results are encouraging.
Problems encountered
On macOS there is a PyTorch issue that hinders multiprocessing for data loading. I couldn't find a proper workaround (see https://stackoverflow.com/questions/64772335/pytorch-w-parallelnative-cpp206). Interestingly, that issue does not appear for regular training (with a local coco dataset), so there are some side effects due to the hub usage.