Agent frameworks & infrastructure

Models, fine-tuning & training

2,504 agents · most starred first

Models, fine-tuning & training, page 3 of 26

Qwen-14B-Chat★ 381

transformers safetensors qwen text-generation custom_code zh

CodeLlama-7b-hf★ 380

transformers pytorch safetensors llama text-generation llama-2

SkyCode-AI-CodeX-GPT3★ 379

SkyCode是一个多语言开源编程大模型,采用GPT3模型结构,支持Java, JavaScript, C, C++, Python, Go, shell等多种主流编程语言,并能理解中文注释。模型可以对代码进行补全,拥有强大解题能力,使您从编程中解放出来,专心于解决更重要的问题。| SkyCode is an open source programming model, which adopts the GPT3 model structure. It supports Java, JavaScript, C, C++, Python, Go, shell and other languages, and can understand Chinese comments.

gym-maze★ 377

A basic 2D maze environment where an agent start from the top left corner and try to find its way to the bottom left corner.

RSIHub★ 377

A research framework for principled agent self-improvement under frozen evaluators and declared mutation boundaries, recording verifiable lineage to make it reproducible and auditable.

gpt2client★ 373

✍🏻 gpt2-client: Easy-to-use TensorFlow Wrapper for GPT-2 117M, 345M, 774M, and 1.5B Transformer Models 🤖 📝

agent-jev★ 369

AgentJev-0.6B - a fast 'System One' decision model for AI Agents: feed it any unstructured state (diffs, traces, logs) and structured questions, get calibrated probability distributions back in one ~50ms forward pass. Zero output-token decoding.

l0★ 368

A scalable, end-to-end training pipeline for general-purpose agents

alf★ 367

Agent Learning Framework https://alf.readthedocs.io

AgentRL★ 367

Scaling Agentic Reinforcement Learning with a Multi-Turn, Multi-Task Framework

TimeChamber★ 366

A Massively Parallel Large Scale Self-Play Framework

Agent-FLAN★ 362

[ACL2024 Findings] Agent-FLAN: Designing Data and Methods of Effective Agent Tuning for Large Language Models

Llama-2-13B-chat-GPTQ★ 362

transformers safetensors llama text-generation facebook meta

pith-train★ 360

Compact and Agent-Native MoE Training System

learning-to-communicate-pytorch★ 359

Learning to Communicate with Deep Multi-Agent Reinforcement Learning in PyTorch

Yi-34B-Chat★ 358

transformers safetensors llama text-generation conversational arxiv:2403.04652

Stable-Alignment★ 355

Multi-agent Social Simulation + Efficient, Effective, and Stable alternative of RLHF. Code for the paper "Training Socially Aligned Language Models in Simulated Human Society".

dqn-multi-agent-rl★ 352

Deep Q-learning (DQN) for Multi-agent Reinforcement Learning (RL)

world-model-optimizer★ 351

Build continually improving models on your agent traces by distilling frontier open models

Step-3.7-Flash★ 348

A high-efficiency Flash model for real-world agents.

gemma-4-31B-it-assistant★ 347

transformers safetensors gemma4_assistant text-generation any-to-any arxiv:2607.02770

Llama3-ChatQA-1.5-70B★ 334

transformers safetensors llama text-generation nvidia chatqa-1.5

openai_lab★ 327

An experimentation framework for Reinforcement Learning using OpenAI Gym, Tensorflow, and Keras.

qqr★ 326

qqr is an RL training framework for open-ended agents.

NexusRaven★ 324

NexusRaven-13B, a new SOTA Open-Source LLM for function calling. This repo contains everything for reproducing our evaluation on NexusRaven-13B and baselines.

RL_Matrix★ 312

Deep Reinforcement Learning in C#

cevahir-ai★ 312

Cevahir Zihin: a mind that learns by living, not a Transformer. It keeps knowledge as evidence on ties, attends with a Seven-Place Focus and learns from every conversation. Also: a Turkish BPE tokenizer, a Transformer language-model core and cognitive systems.

OpenThoughts-Agent★ 312

Data recipes and robust infrastructure for training AI agents

mini-verl★ 312

verl for a single consumer GPU. PPO, GRPO and on-policy distillation on NVIDIA GPUs.

RLHF-V★ 311

[CVPR'24] RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedback

reevo★ 311

[NeurIPS 2024] ReEvo: Large Language Models as Hyper-Heuristics with Reflective Evolution

Libra★ 311

Libra: Efficient Resource Management for Agentic RL Post-Training

NeuralDialog-CVAE★ 310

Tensorflow Implementation of Knowledge-Guided CVAE for dialog generation ACL 2017. It is released by Tiancheng Zhao (Tony) from Dialog Research Center, LTI, CMU

stable-baselines★ 308

Mirror of Stable-Baselines: a fork of OpenAI Baselines, implementations of reinforcement learning algorithms

text2text★ 305

Text2Text Language Modeling Toolkit

openJev-verdict-2.0★ 305

Calibrated 151M Non-Autoregressive Decision Engine beating TypeSafe Jev & Laya on LocalLLaMA/typed-decisions (77.10% acc, 0.0636 Brier, 0.0144 ECE)

RLcycle★ 301

A library for ready-made reinforcement learning agents and reusable components for neat prototyping

FireAct★ 296

FireAct: Toward Language Agent Fine-tuning

Athene-V2-Chat★ 292

transformers safetensors qwen2 text-generation RLHF Nexusflow

kerlym★ 291

KEras Reinforcement Learning gYM agents

unsloth-buddy★ 282

Zero-friction LLM fine-tuning skill for Claude Code, Gemini CLI & any ACP agent. Unsloth on NVIDIA · TRL+MPS/MLX on Apple Silicon. Automates env setup, LoRA training (SFT, DPO, GRPO, vision), post-hoc GRPO log diagnostics, evaluation, and export end-to-end. Part of the Gaslamp AI platform.

agent-distillation★ 282

Official Code Repository for the paper "Distilling LLM Agent into Small Models with Retrieval and Code Tools"

Yi-1.5-34B-Chat★ 280

transformers safetensors llama text-generation conversational arxiv:2403.04652

CLIP-fine-tune★ 274

Fine-tuning code for CLIP models

HarborRL★ 274

An end-to-end framework for Agentic Reinforcement Learning.

sweet_rl★ 273

Benchmark and research code for the paper SWEET-RL Training Multi-Turn LLM Agents onCollaborative Reasoning Tasks

marley★ 272

A framework for multi-agent reinforcement learning.

agentic-grpo-longhorizon★ 272

Fixing GRPO training collapse in long-horizon multi-tool agents. A lightweight PRM-Lite + LATA joint approach achieves +37% over vanilla GRPO on τ-bench airline (50-task, multi-turn).

gpt-2-tensorflow2.0★ 268

OpenAI GPT2 pre-training and sequence prediction implementation in Tensorflow 2.0

Llama-2-7B-Chat-GPTQ★ 267

transformers safetensors llama text-generation facebook meta

Llama3.1-8B-Chinese-Chat★ 266

transformers safetensors gguf llama text-generation llama-factory

SharpToken★ 264

SharpToken is a C# library for tokenizing natural language text. It's based on the tiktoken Python library and designed to be fast and accurate.

Llama-2-70B-Chat-GPTQ★ 259

transformers safetensors llama text-generation facebook meta

Qwen-AgentWorld-35B-A3B-GGUF★ 257

transformers gguf qwen unsloth world-model agent

chatglm3-finetune★ 252

最容易上手的0门槛 chatglm3 & agent & langchain 项目

AgenticArXiv-RL★ 247

Staged SFT→DPO→GRPO/OPD reinforcement learning pipeline for training a ReAct tool-use agent on arXiv paper retrieval tasks, with a deterministic snapshot-replay environment.

deepseek-llm-7b-chat★ 246

transformers pytorch llama text-generation conversational license:other

J-Wash★ 245

Jacobian-Brainwash : A manual alignment tool for large language models built on Anthropic's Jacobian Lens. Results are exportable.

AgentJet★ 244

Cutting-edge platform for LLM agent tuning. Deliver RL tuning with flexibility, reliability, speed, multi-agent optimization and realtime community benchmarking.

ai-trains-ai★ 240

RL-training an AI agent to RL-train AI agents.

SAfactory★ 235

SAfactory: A Scalable Agentic Infrastructure for Training Trustworthy Autonomous Intelligence

debug_seq2seq★ 233

[unmaintained] Make seq2seq for keras work

Qwen3.8-27B-Humanlike-Chat-GGUF★ 233

gguf llama.cpp qwen3.8 conversational roleplay creative-writing

PettingLLMs★ 229

[ICLR'26] Stronger-MAS: A RL Framework for multi LLM agent system; [arxiv] MetaAgent-X: End-to-End Reinforcement Learning Automatic Multi-Agent Systems

reptile-pytorch★ 219

A PyTorch implementation of OpenAI's REPTILE algorithm

empower-functions★ 216

GPT-4 level function calling models for real-world tool using use cases

Tokenizer★ 214

Typescript and .NET implementation of BPE tokenizer for OpenAI LLMs.

autoresearch-qwen★ 212

Autonomous Qwen3-VL training-code research on the official DocVQA benchmark. main: NVIDIA multi-GPU, mlx: Apple Silicon/MPS.

Qwen3-Coder-Next-FP8★ 212

transformers safetensors qwen3_next text-generation conversational license:apache-2.0

RSI-Qwen★ 212

Autonomous Qwen3-VL training-code research on the official DocVQA benchmark. main: NVIDIA multi-GPU, mlx: Apple Silicon/MPS.

spiral★ 211

SPIRAL: Self-Play on Zero-Sum Games Incentivizes Reasoning via Multi-Agent Multi-Turn Reinforcement Learning

Multi-Agent-Reinforcement-Learning★ 210

PyTorch implementations of MADDPG, MAPPO (coming)

ArCHer★ 210

Research Code for "ArCHer: Training Language Model Agents via Hierarchical Multi-Turn RL"

GeDi★ 208

GeDi: Generative Discriminator Guided Sequence Generation

ReConcile★ 208
ToolBrain★ 207

A framework for agentic tool use training with reinforcement learning

nlp-gym★ 203

NLPGym - A toolkit to develop RL agents to solve NLP tasks.

Qwen2.5-Coder-7B★ 202

transformers safetensors qwen2 text-generation code qwen

TaskCraft★ 201

[ICLR 2026] A library for generating difficulty-scalable, multi-tool, and verifiable agentic tasks with execution trajectories.

deepseek-coder-7b-instruct-v1.5★ 195

transformers safetensors llama text-generation conversational license:other

Qwen2.5-Coder-32B★ 191

transformers safetensors qwen2 text-generation code qwen

Deep-Reinforcement-Learning-for-Dialogue-Generation-in-tensorflow★ 189

Deep-Reinforcement-Learning-for-Dialogue-Generation-in-tensorflow

pytorch-sentiment-neuron★ 188
jat★ 187

General multi-task deep RL Agent

Strand-Rust-Coder-14B-v1★ 187

transformers safetensors qwen2 text-generation conversational dataset:Fortytwo-Network/Strandset-Rust-v1

Llama-3-Groq-70B-Tool-Use★ 185

transformers safetensors llama text-generation facebook meta

m-courtyard★ 182

M-Courtyard: Local AI Model Fine-tuning Assistant for Apple Silicon. Zero-code, zero-cloud, privacy-first desktop app powered by Tauri + React + mlx-lm.

raft★ 178

RAFT, or Retrieval-Augmented Fine-Tuning, is a method comprising of a fine-tuning and a RAG-based retrieval phase. It is particularly suited for the creation of agents that realistically emulate a specific human target.

AutoTTS★ 178

The offical repo for "LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling"

rapidfireai★ 177

RapidFire AI: Rapid AI Customization from RAG to Fine-Tuning

deepseek-moe-16b-chat★ 177

transformers safetensors deepseek text-generation conversational custom_code

DrMAS★ 175

Dr. MAS is an end-to-end RL training framework for multi-agent LLM systems, supporting the co-training of multiple (heterogeneous) LLMs.

chatgpt-finetune-ui★ 173

Simple python WebUI for fine-tuning ChatGPT (gpt-3.5-turbo)

DeepSeek-V2-Lite-Chat★ 173

transformers safetensors deepseek_v2 text-generation conversational custom_code

Qwen2.5-Coder-1.5B-Instruct★ 173

transformers safetensors qwen2 text-generation code codeqwen

Qwen3-Coder-480B-A35B-Instruct-FP8★ 171

transformers safetensors qwen3_moe text-generation conversational arxiv:2505.09388

ResearchBoost-Agent★ 171

An autonomous agent that profiles, optimizes, and validates AI training code.

Qwen2.5-Coder-1.5B-Instruct-GGUF★ 170

transformers gguf code codeqwen chat qwen

jais-13b-chat★ 168

transformers pytorch jais text-generation Arabic English

Qwen-72B-Chat★ 164

transformers safetensors qwen text-generation custom_code zh

More in Agent frameworks & infrastructure