transformers
feat(trainer): Just-in-time (JIT) asynchronous checkpointing using SIGTERM signals
#41723

Merged

feat(trainer): Just-in-time (JIT) asynchronous checkpointing using SIGTERM signals #41723

SunMarc merged 8 commits into huggingface:main from efazal:feat-jit-checkpointing

efazal force pushed from a457b744 to 05411f37 69 days ago

stas00 commented on 2025-10-22

efazal force pushed from e093fcd7 to 5b21304d 63 days ago

SunMarc requested a review from

ArthurZucker 56 days ago

SunMarc requested a review from

SunMarc 56 days ago

SunMarc requested a review from

qgallouedec 54 days ago

efazal force pushed from 5b21304d to 16b0a402 43 days ago

SunMarc commented on 2025-11-19

efazal force pushed from 16b0a402 to 00a9c2b6 30 days ago

Just-in-time (JIT) asynchronous checkpointing using SIGTERM signals a…

b95a867b

Fix failing ci tests

a1389622

Update JIT checkpoint code to remove CUDA streams and async checkpoin…

44433fb0

Fix sentinel file save path to checkpoint folder, update checkpoint r…

4ab2427d

Refactor JIT checkpoint logic: rename methods and variables for clari…

929d2dcc

Remove support for environment variable overrides in `TrainingArgumen…

1eb7f1f2

efazal force pushed from 434c4b58 to 1eb7f1f2 29 days ago

Merge branch 'main' into feat-jit-checkpointing

6e6f4c26

Apply style fixes

57a9df16

SunMarc approved these changes on 2025-12-03

SunMarc merged fda2d735 into main 26 days ago

Reviewers

SunMarc

stas00

ArthurZucker

qgallouedec

Assignees

No one assigned

Labels

None yet

Milestone

No milestone