423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-23 06:12:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

lint

Browse Source 
Former-commit-id: da9e4ddd26ebd6e7eb266aa0bef7505465a6b119
This commit is contained in:
hoshi-hiyouga 2024-11-25 22:55:56 +08:00 committed by GitHub
parent a6aeb98af6
commit 118ffe50e3
1 changed files with 8 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
src/llamafactory/data/mm_plugin.py
View File

@ -753,12 +753,14 @@ class MllamaPlugin(BasePlugin):
cross_attention_token_mask = [
get_cross_attention_token_mask(input_ids, image_token_id) for input_ids in batch_ids
]
mm_inputs["cross_attention_mask"] = torch.tensor(convert_sparse_cross_attention_mask_to_dense(
cross_attention_token_mask,
num_tiles=num_tiles,
max_num_tiles=max_image_tiles,
length=max(len(input_ids) for input_ids in batch_ids),
))
mm_inputs["cross_attention_mask"] = torch.from_numpy(
convert_sparse_cross_attention_mask_to_dense(
cross_attention_token_mask,
num_tiles=num_tiles,
max_num_tiles=max_image_tiles,
length=max(len(input_ids) for input_ids in batch_ids),
)
)
return mm_inputs