This repository has been archived by the owner on Nov 3, 2023. It is now read-only.

Now multiple checkpoints will be saved after using -stopk option in train_model.py#4978

Open

hamjam wants to merge 13 commits intofacebookresearch:mainfrom hamjam:multiple_checkpoint

+81-7

Commits on Mar 8, 2023

Multiple checkpoint save
Hamed Jamshidian
committed
Fixing duplicate topk opt
Hamed Jamshidian
committed
Fixing not casting to string _train_steps in model_suffix
Hamed Jamshidian
committed
Reseting impatience in save_top_k mode
Hamed Jamshidian
committed
Fixing bug in suffix ordinal
Hamed Jamshidian
committed

Commits on Mar 10, 2023

Fixing bug in _modify_next_rank_checkpoints ordinal
Hamed Jamshidian
committed
Fixed bug in indexing checkpoints after the new checkpoint
Hamed Jamshidian
committed
Using pathlib to remove redundant files in multiple checkpoint stats
Hamed Jamshidian
committed
Logging for ranning out of patience in multiple checkpoints state
Hamed Jamshidian
committed
Storing save state in multiple checkpoint
Hamed Jamshidian
committed
Stop training in multiple checkpoints if validation metric passed cutoff
Hamed Jamshidian
committed
Merge branch 'facebookresearch:main' into multiple_checkpoint
hamjam
authored

Commits on Sep 16, 2023

Merge branch 'facebookresearch:main' into multiple_checkpoint
hamjam
authored