This repository was archived by the owner on Jul 30, 2026. 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 into

facebookresearch:mainfrom

hamjam:multiple_checkpoint

Commits on Mar 8, 2023

Multiple checkpoint save
Hamed Jamshidian
committed
8db245a
Fixing duplicate topk opt
Hamed Jamshidian
committed
8edc746
Fixing not casting to string _train_steps in model_suffix
Hamed Jamshidian
committed
d40adef
Reseting impatience in save_top_k mode
Hamed Jamshidian
committed
4a066d6
Fixing bug in suffix ordinal
Hamed Jamshidian
committed
c7e1dc0

Commits on Mar 10, 2023

Fixing bug in _modify_next_rank_checkpoints ordinal
Hamed Jamshidian
committed
373d0a1
Fixed bug in indexing checkpoints after the new checkpoint
Hamed Jamshidian
committed
a5a2ab5
Using pathlib to remove redundant files in multiple checkpoint stats
Hamed Jamshidian
committed
e01b6de
Logging for ranning out of patience in multiple checkpoints state
Hamed Jamshidian
committed
dac1c81
Storing save state in multiple checkpoint
Hamed Jamshidian
committed
3918b8f
Stop training in multiple checkpoints if validation metric passed cutoff
Hamed Jamshidian
committed
a0da2f0
Merge branch 'facebookresearch:main' into multiple_checkpoint
hamjam
authored
21a5baa

Commits on Sep 16, 2023

Merge branch 'facebookresearch:main' into multiple_checkpoint
hamjam
authored
6dc9f98