Scripts for fine-tuning Meta Llama with composable FSDP & PEFT methods to cover single/multi-node GPUs. Supports default & custom datasets for applications such as summarization and Q&A. Supporting a number of candid inference solutions such as HF TGI, VLLM for local or cloud deployment. Demo apps to showcase Meta Llama for WhatsApp & Messenger.

Replace OpenAI GPT with another LLM in your app by changing a single line of code. Xinference gives you the freedom to use any LLM you need. With Xinference, you're empowered to run inference with any open-source language models, speech recognition models, and multimodal models, whether in the cloud, on-premises, or even on your laptop.

Data processing with ML, LLM and Vision LLM

📖A curated list of Awesome LLM Inference Paper with codes, TensorRT-LLM, vLLM, streaming-llm, AWQ, SmoothQuant, WINT8/4, Continuous Batching, FlashAttention, PagedAttention etc.

An Easy-to-use, Scalable and High-performance RLHF Framework (70B+ PPO Full Tuning & Iterative DPO & LoRA & RingAttention)

🔒 Enterprise-grade API gateway that helps you monitor and impose cost or rate limits per API key. Get fine-grained access control and monitoring per user, application, or environment. Supports OpenAI, Azure OpenAI, Anthropic, vLLM, and open-source LLMs.

Evaluate your LLM's response with Prometheus and GPT4 💯

Private Open AI on Kubernetes

Low latency JSON generation using LLMs ⚡️

[EMNLP 2024 Industry Track] This is the official PyTorch implementation of "LLMC: Benchmarking Large Language Model Quantization with a Versatile Compression Toolkit".

A large-scale simulation framework for LLM inference

The goal of RamaLama is to make working with AI boring.

The RunPod worker template for serving our large language model endpoints. Powered by vLLM.

Official Implementation of TopicGPT: A Prompt-Based Framework for Topic Modeling (NAACL '24)

Setup and run a local LLM and Chatbot using consumer grade hardware.

【深度学习模型部署框架】支持tf/torch/trt/trtllm/vllm以及更多nn框架，支持dynamic batching、streaming模式，支持python/c++双语言，可限制，可拓展，高性能。帮助用户快速地将模型部署到线上，并通过http/rpc接口方式提供服务。

Documentation on setting up an LLM server on Debian from scratch, using Ollama/vLLM, Open WebUI, OpenedAI Speech, and ComfyUI.

Booster - open accelerator for LLM models. Better inference and debugging for AI hackers

Official implementation of "DoRA: Weight-Decomposed Low-Rank Adaptation"

Framework agnostic computer vision inference. Run 1000+ models by changing only one line of code. Supports models from transformers, timm, ultralytics, vllm, ollama and your custom model.