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

fix: apply query RoPE when reusing cached vision KV; align processor with the 0708 standard

Align tokenizer_config with the standard (embed chat_template)

Upload README.md with huggingface_hub

Fix language switch links

Separate English and Chinese benchmark model cards

Add bilingual quantization benchmark comparison

Restore model card logo

Add files using upload-large-folder tool

initial commit