Work in progress...
Data is assumed to be the iNaturalist dataset for the convolutional benchmark.
Most of the code here follows PyTorch examples.
For dev purposes:
python inaturalist.py ./data --workers 8 -b 12 --gpu 0
For filling up four RTX 2080Ti
and training with torch.DataParallel
:
python inaturalist.py ./data -a resnet50 --workers 32 -b 192
- Write a proper Readme.
- Add dependencies.
- Read from
yaml
config files the standardized benchmarks. - Start adding additional models.