LMDeploy is a toolkit for compressing, deploying, and serving LLMs.

Safe RLHF: Constrained Value Alignment via Safe Reinforcement Learning from Human Feedback

An Open-sourced Knowledgable Large Language Model Framework.

Best practice for training LLaMA models in Megatron-LM

Guide: Finetune GPT2-XL (1.5 Billion Parameters) and finetune GPT-NEO (2.7 B) on a single GPU with Huggingface Transformers using DeepSpeed

Collaborative Training of Large Language Models in an Efficient Way

GLake: optimizing GPU memory management and IO transmission.

Best practices & guides on how to write distributed pytorch training code

Implementation of Chinese ChatGPT

一个基于HuggingFace开发的大语言模型训练、测试工具。支持各模型的webui、终端预测，低参数量及全参数模型训练(预训练、SFT、RM、PPO、DPO)和融合、量化。

llama2 finetuning with deepspeed and lora

Super-Efficient RLHF Training of LLMs with Parameter Reallocation

DeepSpeed教程 & 示例注释 & 学习笔记 （大模型高效训练）

A full pipeline to finetune ChatGLM LLM with LoRA and RLHF on consumer hardware. Implementation of RLHF (Reinforcement Learning with Human Feedback) on top of the ChatGLM architecture. Basically ChatGPT but with ChatGLM

Simple and efficient RevNet-Library for PyTorch with XLA and DeepSpeed support and parameter offload

Train llm (bloom, llama, baichuan2-7b, chatglm3-6b) with deepspeed pipeline mode. Faster than zero/zero++/fsdp.

llm-inference is a platform for publishing and managing llm inference, providing a wide range of out-of-the-box features for model deployment, such as UI, RESTful API, auto-scaling, computing resource management, monitoring, and more.

Scripts of LLM pre-training and fine-tuning (w/wo LoRA, DeepSpeed)

A full pipeline to finetune Alpaca LLM with LoRA and RLHF on consumer hardware. Implementation of RLHF (Reinforcement Learning with Human Feedback) on top of the Alpaca architecture. Basically ChatGPT but with Alpaca

A toy large model for recommender system based on LLaMA2/SASRec/Meta's generative recommenders. Besides, note and experiments of official implementation for Meta's generative recommenders.