Types & classes — github.com/Ironieser/MMTok

CodeHub MCPcopy Create free account

hub / github.com/Ironieser/MMTok / types & classes

Types & classes8 in github.com/Ironieser/MMTok

⨍ Functions 75 ◇ Types & classes 8

↓ 2 callersClassMMTokCore

MMTok core: coverage-based subset selection of vision tokens. Uses both vision and text (question/keywords) to select informative vision

mmtok/core/mmtok_core.py:31

↓ 1 callersClassSemanticTokenSelector

Coverage-based subset selection of vision tokens. Greedy selection to cover text tokens (P) and the vision token set (Q); score = covera

mmtok/core/semantic_selector.py:68

↓ 1 callersClassVQATextProcessor

Text processor for MMTok: extract question/keywords used as text tokens in the coverage criterion (cover text tokens + vision token set).

mmtok/core/text_processor.py:34

ClassCLIPVisionTower_MMTok

MMTok vision tower: CLIP -> mm_projector -> subset selection (coverage criterion).

mmtok/llava/clip_encoder_mmtok.py:21

ClassLlava_MMTok

Llava Model with MMTok.

example/lmms_eval_llava_mmtok.py:51

ClassQwen2_5_VL_MMTok

Qwen2.5_VL with MMTok.

example/lmms_eval_qwen_mmtok.py:33

ClassQwen2_5_VL_MMTok

MMTok Qwen2.5-VL: forward runs vision -> MMTok selection -> language model.

mmtok/qwen/qwen2_5_VLmodel_mmtok.py:19

ClassQwen2_5_VisionTransformerPretrainedModel_MMTok

mmtok/qwen/modeling_qwen2_5_vl_mmtok.py:13