Kaffaljidhmah2

gpt-fast

Simple and efficient pytorch-native transformer text generation in <1000 LOC of python.

SpeculativeDecodingPapers

📰 Must-read papers and blogs on Speculative Decoding ⚡️

smart_router

A smart router to switch between GPT-3.5 and GPT-4 based on the hardness of the context. Aim to reduce cost while keeping the performance ≈ GPT-3¾.

Organized-LLM-Agents

Embodied and organized multi-LLM-agent teams supporting communication for >3 agents. Source codes for the paper "Embodied LLM Agents Learn to Cooperate in Organized Teams".

c4-dataset-script

Inspired by google c4, here is a series of colossal clean data cleaning scripts focused on CommonCrawl data processing. Including Chinese data processing and cleaning methods in MassiveText.

gemma

Open weights LLM from Google DeepMind.

grok-1

Grok open release

awesome-pydantic

A curated list of awesome things related to Pydantic! 🌪️

gpt_academic

为GPT/GLM等LLM大语言模型提供实用化交互接口，特别优化论文阅读/润色/写作体验，模块化设计，支持自定义快捷按钮&函数插件，支持Python和C++等项目剖析&自译解功能，PDF/LaTex论文翻译&总结功能，支持并行问询多种LLM模型，支持chatglm3等本地模型。接入通义千问, deepseekcoder, 讯飞星火, 文心一言, llama2, rwkv, claude2, moss等。

instructor

structured outputs for llms

REST

REST: Retrieval-Based Speculative Decoding, NAACL 2024

langfun

Empower LLMs with Symbols.

LoRA-EXTRACTOR-Colab

A small script to extract LoRA models from custom checkpoints, in Google Colab.

ToRA

ToRA is a series of Tool-integrated Reasoning LLM Agents designed to solve challenging mathematical reasoning problems by interacting with tools [ICLR'24].

LLM-Shearing

[ICLR 2024] Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning

Co-LLM-Agents

Source codes for the paper "Building Cooperative Embodied Agents Modularly with Large Language Models"

grace

[EMNLP 2023, Findings] GRACE: Discriminator-Guided Chain-of-Thought Reasoning

LLMSpeculativeSampling

Fast inference from large lauguage models via speculative decoding

gorilla

Gorilla: An API store for LLMs

Jailbreak_LLM

TheoremQA

The dataset and code for paper: TheoremQA: A Theorem-driven Question Answering dataset

arxiv-gpt

An extension of chaotic_neural to visualize papers clustered using GPT-based embeddings

llm-awq

[MLSys 2024 Best Paper Award] AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration

MLAgentBench

Medusa

Medusa: Simple Framework for Accelerating LLM Generation with Multiple Decoding Heads

FlexFlow

FlexFlow Serve: Low-Latency, High-Performance LLM Serving

UltraFeedback

A large-scale, fine-grained, diverse preference dataset (and models).

RL4LMs

A modular RL library to fine-tune language models to human preferences

trl

Train transformer language models with reinforcement learning.

alpaca_farm

A simulation framework for RLHF and alternatives. Develop your RLHF method without collecting human data.