XTuner is a toolkit for efficiently fine-tuning LLM

🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

verl: Volcano Engine Reinforcement Learning for LLMs

SGLang is a fast serving framework for large language models and vision language models.

A Scientific Multimodal Foundation Model

slime is a LLM post-training framework aiming at scaling RL.

Unified Efficient Fine-Tuning of 100+ LLMs (ACL 2024)

Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 500+ LLMs (Qwen3, Qwen3-MoE, Llama4, GLM4.5, InternLM3, DeepSeek-R1, ...) and 200+ MLLMs (Qwen2.5-VL, Qwen2.5-Omni, Qwen2-Audio, InternVL3, Ovis2.5, L…

Towards Economical Inference: Enabling DeepSeek's Multi-Head Latent Attention in Any Transformer-based LLMs

A native PyTorch Library for large model training

Example models using DeepSpeed

DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

Efficient Triton Kernels for LLM Training

Janus-Series: Unified Multimodal Understanding and Generation Models

USP: Unified (a.k.a. Hybrid, 2D) Sequence Parallel Attention for Long Context Transformers Model Training and Inference

Open-source evaluation toolkit of large vision-language models (LVLMs), support ~100 VLMs, 40+ benchmarks

Train transformer language models with reinforcement learning.

[CVPR 2024 Oral] InternVL Family: A Pioneering Open-Source Alternative to GPT-4o. 接近GPT-4o表现的开源多模态对话模型

Ring attention implementation with flash attention

A GPipe implementation in PyTorch

The official Meta Llama 3 GitHub site

Accelerating the development of large multimodal models (LMMs) with lmms-eval

OpenMMLab Detection Toolbox and Benchmark

Reformatted Alignment

Course to get into Large Language Models (LLMs) with roadmaps and Colab notebooks.

Project Page for "LISA: Reasoning Segmentation via Large Language Model"

Visual Instruction Tuning: Large Language-and-Vision Assistant built towards multimodal GPT-4 level capabilities.