Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 500+ LLMs (Qwen3, Qwen3-MoE, Llama4, InternLM3, DeepSeek-R1, ...) and 200+ MLLMs (Qwen2.5-VL, Qwen2.5-Omni, Qwen2-Audio, Ovis2, InternVL3, Llava, GLM4v, Phi4, ...) (AAAI 2025).
deploy llama lora embedding omni liger peft multimodal sft megatron rft llm internvl qwen2-vl deepseek-r1 grpo open-r1 qwen3 llama4 qwen3-moe
-
Updated
Jun 18, 2025 - Python