[megatron] fix: critic and reward model load tokenizer from config #301
+20
−3
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
HollowMan6
force-pushed
the
tokenizer
branch
2 times, most recently
from
9c51e60 to
ab24a03
Compare
|
Just fixed the yapf formatting issue |
HollowMan6
force-pushed
the
tokenizer
branch
from
ab24a03 to
9b45db2
Compare
Currently, the worker will fail if the critic or reward model path doesn't contain a tokenizer. This PR tries to fix this by loading tokenizer from the config for the previously mentioned case. - For the critic model, we fall back to load from `critic.model.tokenizer_path`. - For the reward model, we first fall back to load from `reward_model.model.rm_tokenizer`, and then `reward_model.model.input_tokenizer` if that is not set. Signed-off-by: Hollow Man <[email protected]>
HollowMan6
force-pushed
the
tokenizer
branch
from
9b45db2 to
2864034
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Currently, the worker will fail if the critic or reward model path doesn't contain a tokenizer. This PR tries to fix this by loading tokenizer from the config for the previously mentioned case.
critic.model.tokenizer_path.reward_model.model.rm_tokenizer, and thenreward_model.model.input_tokenizerif that is not set.