Agent: Cursor, Claude CodeLLM: Qwen, Claude#llm-training#fine-tuning#multimodal#reinforcement-learning#open-source
MS-Swift is a scalable framework for fine-tuning and deploying large language models and multimodal models. It supports 600+ LLMs (Qwen, DeepSeek, Llama, GLM, InternLM) and 300+ MLLMs with training methods including CPT, SFT, DPO, and GRPO. Includes distributed training via Megatron, quantization, evaluation, and deployment capabilities.