423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-04 12:42:51 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix baichuan resize

Browse Source 
Former-commit-id: f4657de7d574fdab5d164c679bd474d35140894a
This commit is contained in:
hiyouga 2023-12-11 20:55:50 +08:00
parent 2e42e38ff2
commit e3e86340ec
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/llmtuner/model/utils.py
View File

@ -188,6 +188,10 @@ def resize_embedding_layer(model: "PreTrainedModel", tokenizer: "PreTrainedToken
r""" r"""
Resize token embeddings. Resize token embeddings.
""" """
if not isinstance(model.get_output_embeddings(), torch.nn.Linear):
logger.warning("Current model does not support resizing token embeddings.")
return
old_vocab_size = model.get_input_embeddings().weight.size(0) old_vocab_size = model.get_input_embeddings().weight.size(0)
if len(tokenizer) != old_vocab_size: if len(tokenizer) != old_vocab_size:
model.resize_token_embeddings(len(tokenizer), pad_to_multiple_of=64) model.resize_token_embeddings(len(tokenizer), pad_to_multiple_of=64)