refactor: minor refactors regarding multimodal processing (#6187)

This commit is contained in:
Mick
2025-05-17 22:53:20 -07:00
committed by GitHub
parent b3f3d610fd
commit 01dd39bac1
15 changed files with 140 additions and 98 deletions
+5 -2
View File
@@ -1520,12 +1520,15 @@ class MiniCPMO(MiniCPMBaseModel):
slice_start_id: int = mm_input.slice_start_id
slice_end_id: int = mm_input.slice_end_id
media_token_pairs = [
data_token_pairs = [
(im_start_id, im_end_id),
(slice_start_id, slice_end_id),
(mm_input.audio_start_id, mm_input.audio_end_id),
]
pattern = MultiModalityDataPaddingPatternTokenPairs(media_token_pairs)
data_start_token_ids = [im_start_id, mm_input.audio_start_id]
pattern = MultiModalityDataPaddingPatternTokenPairs(
data_token_pairs=data_token_pairs, data_start_token_ids=data_start_token_ids
)
return pattern.pad_input_tokens(input_ids, mm_input)