Skip to content

allow manual token_ids to be passed to fix untrained lrs #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Dec 23, 2024
Merged

allow manual token_ids to be passed to fix untrained lrs #2

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f7516be
allow manual token_ids to be passed to fix untrained lrs
winglian Dec 22, 2024
2722306
fix set operation
winglian Dec 22, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion src/axolotl/contribs/lgpl/unsloth.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,14 +24,16 @@

@torch.inference_mode()
def fix_untrained_tokens( # pylint: disable=too-many-return-statements
model, tokenizer, train_dataset, ignored_tokenizer_names=None, eps=1e-16
model, tokenizer, train_dataset, ignored_tokenizer_names=None, eps=1e-16, token_ids_to_fix=None,
):
"""
Llama-3 for eg has untrained vectors in the base model.
These include <|eot_id|>, <|start_header_id|>, <|end_header_id|>
We reset them to the mean of the rest of the tokens
"""
# Code licensed under LGPL
if not token_ids_to_fix:
token_ids_to_fix = []
embedding_matrix = model.get_input_embeddings().weight
lm_head_matrix = model.get_output_embeddings().weight
chat_template = getattr(tokenizer, "chat_template", None)
Expand Down Expand Up @@ -91,6 +93,7 @@ def fix_untrained_tokens( # pylint: disable=too-many-return-statements

# Get set and actual tokens
where_untrained = where_untrained.tolist()
where_untrained = list(set(token_ids_to_fix + where_untrained))
if len(where_untrained) == 0:
return

Expand Down