ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update llama.py

Browse source
This commit is contained in:
Daniel Han 2024-07-20 12:47:36 -07:00
commit ba515ec92d
1 changed files with 8 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

9
unsloth/models/llama.py
View file

@ -1068,7 +1068,14 @@ def _wrap_fast_inference(generate, device_type, dtype, model):
# For newer HF
kwargs["cache_implementation"] = "dynamic"
print(kwargs)
# Remove token_type_ids
kwargs.pop("token_type_ids", None)
# Check pad_token
kwargs["pad_token_id"] = kwargs.pop(
"pad_token_id",
getattr(model.config, "eos_token_id", None),
)
# Set pad token
# old_pad_token_id = getattr(model.config, "pad_token_id", None)