ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[misc] fix script (#6977)

Browse Source 
Former-commit-id: 775efa1d8cbdb1b7d122be2a986d47f85214e0a1
This commit is contained in:
hoshi-hiyouga
2025-02-18 17:00:46 +08:00
committed by GitHub
parent f5cd17881e
commit be33ef67fb
3 changed files with 22 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/llamafactory/data/mm_plugin.py
View File

@@ -1112,9 +1112,13 @@ class Qwen2vlPlugin(BasePlugin):
self._validate_input(images, videos, audios)
image_processor: "BaseImageProcessor" = getattr(processor, "image_processor")
merge_length: int = getattr(image_processor, "merge_size") ** 2
mm_inputs = self._get_mm_inputs(images, videos, audios, processor)
image_grid_thw = mm_inputs.get("image_grid_thw", [])
video_grid_thw = mm_inputs.get("video_grid_thw", [])
if self.expand_mm_tokens:
mm_inputs = self._get_mm_inputs(images, videos, audios, processor)
image_grid_thw = mm_inputs.get("image_grid_thw", [])
video_grid_thw = mm_inputs.get("video_grid_thw", [])
else:
image_grid_thw = [None] * len(images)
video_grid_thw = [None] * len(videos)
num_image_tokens, num_video_tokens = 0, 0
messages = deepcopy(messages)