add reference model logps to chunkedloss interface and fix dpo loss fn #405
Conversation
This file contains hidden or 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
…s use reference model
shivam15s
force-pushed
the
shisahni/dpo_ref
branch
from
116ee1a to
15cb383
Compare
shivam15s
changed the title
shivam15s
changed the title
pramodith
reviewed
Nov 21, 2024
| input_chunk, ref_weight, target_chunk, ref_bias=None, ignore_index=-100 | ||
| ): | ||
| with torch.no_grad(): | ||
| ref_logits_chunk = input_chunk @ ref_weight.t() |
There was a problem hiding this comment.
The logic to get the log_probs is common for both the reference and active policy model, wondering if we can create a util function that can be called by both instead of repeating the code.
There was a problem hiding this comment.
should be possible; just need the util fn to compute nll loss for the active policy forward
shivam15s
changed the title
pramodith
approved these changes
Nov 22, 2024
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Outdated suggestions cannot be applied.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.
accomodate reference model logps in chunked loss interface and make dpo loss use reference model logps in its loss function
Summary
as title
Testing Done
make testto ensure correctnessmake checkstyleto ensure code stylemake test-convergenceto ensure convergence