You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

Lorenzo Baraldi e09b4d5c7f Gradient accumulation included into training script ... Added parameter iters_to_accumulate to perform gradient accumulation		2 years ago
..
__init__.py	Refactor device handling in scripts, distributed init to be less 'cuda' centric. More device args passed through where needed.	2 years ago
agc.py	Initial AGC impl. Still testing.	4 years ago
checkpoint_saver.py	Fix #387 so that checkpoint saver works with max history of 1. Add checkpoint-hist arg to train.py.	4 years ago
clip_grad.py	Initial AGC impl. Still testing.	4 years ago
cuda.py	Gradient accumulation included into training script	2 years ago
decay_batch.py	Add bulk_runner script and updates to benchmark.py and validate.py for better error handling in bulk runs (used for benchmark and validation result runs). Improved batch size decay stepping on retry...	2 years ago
distributed.py	Refactor device handling in scripts, distributed init to be less 'cuda' centric. More device args passed through where needed.	2 years ago
jit.py	disable nvfuser for jit te/legacy modes (for PT 1.12+)	2 years ago
log.py	Reorg of utils into separate modules	4 years ago
metrics.py	Tweak accuracy topk safety. Fix #807	3 years ago
misc.py	Reorg of utils into separate modules	4 years ago
model.py	Fix some formatting in utils/model.py	3 years ago
model_ema.py	Add separate set and update method to ModelEmaV2	4 years ago
random.py	Remove commented code, add more consistent seed fn	4 years ago
summary.py	Scheduler update, add v2 factory method, support scheduling on updates instead of just epochs. Add LR to summary csv. Add lr_base scaling calculations to train script. Fix #1168	2 years ago