feat: Adds embed_tokens, lm_head as trainable for vocab expansion in peft and enables tying of adapters #625
+227
−25
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
Signed-off-by: romit <romit@ibm.com>
|
Thanks for making a pull request! 😃 |
romitjain
commented
Oct 30, 2025
Signed-off-by: romit <romit@ibm.com>
romitjain
requested review from
aluu317,
anhuong,
dushyantbehl,
fabianlim and
kmehant
as code owners
romitjain
commented
Nov 4, 2025
romitjain
commented
Nov 4, 2025
pyproject.toml
Outdated
| "tqdm>=4.66.2,<5.0", | ||
| "trl>=0.19.1,<0.20.0", | ||
| "peft @ git+https://github.com/huggingface/peft.git@293aea5df6db240856a77f89955d1a89ce38b50d", | ||
| "peft @ git+https://github.com/romitjain/peft.git@8388aa869473a60589a01e6950ea0583d3612783", |
Contributor
Author
There was a problem hiding this comment.
This is a temporary path to my fork so that the tests can be validated. Once my PR in PEFT is merged, we can point to huggingface peft
Signed-off-by: romit <romit@ibm.com>
Signed-off-by: romit <romit@ibm.com>
romitjain
commented
Nov 7, 2025
Contributor
Author
There was a problem hiding this comment.
I have added copy.deepcopy for PEFT_LORA_ARGS as the trainer was updating this inplace affecting other tests.
romitjain
commented
Nov 7, 2025
| "pad_token_id": 0, | ||
| "rms_norm_eps": 1e-06, | ||
| "tie_word_embeddings": false, | ||
| "tie_word_embeddings": true, |
Contributor
Author
There was a problem hiding this comment.
This was all that was required to use this model in my tests. This does not break any other test (AFAIK)
Signed-off-by: romit <romit@ibm.com>
dushyantbehl
added
the
on hold
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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.
Description of the change
embed_tokensandlm_headas trainable layersmodules_to_save, enable the flag to ensure weight tying between adapters of the tied layerstarget_modules, enable the flag to ensure weight tying between adapters of the tied layersRelated issue number
Closes https://github.ibm.com/ai-foundation/watson-fm-stack-tracker/issues/1673
How to verify the PR
For 1
For 2 and 3
Was the PR tested