verl: Volcano Engine Reinforcement Learning for LLMs

Defeating the Training-Inference Mismatch via FP16

✨ Build a beautiful and simple website in literally minutes. Demo at https://beautifuljekyll.com

Revisiting Mid-training in the Era of Reinforcement Learning Scaling

Async pipelined version of Verl

This is a replicate of DeepSeek-R1-Zero and DeepSeek-R1 training on small models with limited data

Synthetic Data Generation for Evaluation

The TinyLlama project is an open endeavor to pretrain a 1.1B Llama model on 3 trillion tokens.

Deep Learning for Code Website

A high-throughput and memory-efficient inference and serving engine for LLMs

🌾 OAT: Online AlignmenT for LLMs

Go ahead and axolotl questions

DataComp for Language Models

Extract a single expert from an MoE model of Mixtral architecture, using slerp

Data Toolkit for Sailor Language Models

更纯粹、更高压缩率的Tokenizer

This project studies the performance and robustness of language models and task-adaptation methods.

Puzzles for learning Triton

Accurate line-level text detection and recognition (OCR) in any language

Tools for merging pretrained large language models.

OpenAgents: An Open Platform for Language Agents in the Wild

🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.

distributed trainer for LLMs

Fine-tune SantaCoder for Code/Text Generation.