Make checkpoint saving folder clear in the config#444
Make checkpoint saving folder clear in the config#444DNXie merged 1 commit intometa-pytorch:mainfrom
folder clear in the config#444Conversation
Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags:
meta-cla
bot
added
the
CLA Signed
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #444 +/- ##
==========================================
- Coverage 64.69% 64.63% -0.06%
==========================================
Files 79 79
Lines 7775 7788 +13
==========================================
+ Hits 5030 5034 +4
- Misses 2745 2754 +9 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
| enable: true | ||
| initial_load_path: hf://${model} | ||
| initial_load_in_hf: true | ||
| folder: ./checkpoint # The folder to save checkpoints to. |
There was a problem hiding this comment.
I think we control these config fields and we should be opinionated on exposing RL friendly config field names and re-map it to TorchTitan fields internally.
Right now, some TorchTitan config names are really confusing: e.g. the ref model logically should not need checkpointing but still it requires checkpoint.enable = true.
There was a problem hiding this comment.
Yeah was discussing this with @joecummings too.. ultimately we probably want to provide some kind of internal config mapping that we execute as a kind of training script post-init before we do any of the actual setup of actors (e.g. we can even bake it into our config.parse decorator).
4 participants
Made it clear in the config where to specify the checkpoint saving folder. And added some necessary comments to clarify the behavior.