per-sample media budgets in processor; per-sample MRoPE position ids for batched vision inputs

per-sample media budgets in processor; per-sample MRoPE position ids for batched vision inputs

sync offline system prompts with training data: no_thinking -> no system message, deep_thinking -> <think>/<answer> tags

Unify modeling/configuration/tokenizer with the 0708 standard (transformers 4.57 compatible, includes query RoPE fix)

fix: apply query RoPE in cross-attention when reusing cached vision KV during decode

Replace processor with 0708 version (Fast image/video processor, transformers 4.57+); keep 0408 pixel caps

Upload README.md with huggingface_hub

Update README.md

Update modeling_moss_vl.py

Update video_processing_moss_vl.py

Update processing_moss_vl.py

update transformers to 5.5.4

Update code for transformers 5.5.4 (#7)

Update code for transformers 5.5.4 (#6)

Update code for transformers 5.5.4 (#5)

Update code for transformers 5.5.4 (#4)

Update code for transformers 5.5.4 (#3)

Update code for transformers 5.5.4 (#2)

Update code for transformers 5.5.4 (#1)

Delete assets/MOSS-VL-benchmark.png

Update README.md

Update README.md

Update README.md

Update README.md

Update README.md

Update README.md

Upload structure.png

Delete assets/structure.png

Update README.md

Update README.md

Update README.md

Update README.md

Delete logo.png

Upload logo.png

Upload logo.png

Delete assets/logo.png

Update README.md

Update README.md

Update README.md

Update README.md

Upload 3d-rope.png

Delete assets/3d-rope.png

Update README.md

Update README.md

Update README.md

Update modeling_moss_vl.py

Upload 3d-rope.png

Delete assets/3d-rope.png

Update README.md

Update README.md