Open-source llm-training software
812 open-source libraries, tools and applications tagged llm-training, ranked by how closely llm-training is their primary purpose, then by adoption and maintenance health. Each entry links to a page with sources, use cases and when to choose or avoid it.
812 products, primary matches first, then adoption-weighted; health v2 shown.
| Product | Health v2 | Stars | Maturity |
|---|---|---|---|
| modelscope/AgentEvolver AgentEvolver is an end-to-end self-evolving training framework for LLM-based agents that unifies automatic task generation (self-questionin… | 47 | 1571 | active |
| decoderesearch/SAELens SAELens is a Python library for training sparse autoencoders (SAEs) on language model activations and analyzing them for mechanistic interp… | 90 | 1555 | active |
| yuantianyuan01/FastWAM Official PyTorch codebase for Fast-WAM, a World Action Model for robot manipulation that skips test-time future video imagination and gener… | 56 | 1547 | active |
| MoonshotAI/Moonlight Moonlight is a 3B/16B Mixture-of-Experts LLM trained with the Muon optimizer, released by Moonshot AI along with a memory- and communicatio… | 33 | 1546 | active |
| RLHFlow/RLHF-Reward-Modeling A collection of training recipes for reward models used in RLHF, covering Bradley-Terry reward models, pairwise preference models, ArmoRM, … | 30 | 1546 | active |
| baichuan-inc/Baichuan-7B Baichuan-7B is an open-source, commercially usable 7-billion-parameter pretrained language model built on the Transformer architecture, tra… | 30 | 5648 | maintenance |
| ATH-MaaS/Marco-o1 Marco-o1 is an open large reasoning model from Alibaba International Digital Commerce, designed for o1-like chain-of-thought reasoning acro… | 59 | 1537 | active |
| allenzren/open-pi-zero An open-source re-implementation of the pi0 vision-language-action (VLA) model from Physical Intelligence, built on a pre-trained PaliGemma… | 24 | 1533 | active |
| open-thought/reasoning-gym Reasoning Gym is a Python library of procedural dataset generators and algorithmically verifiable reasoning environments for training LLMs … | 60 | 1512 | active |
| imoneoi/openchat OpenChat is a library of open-source large language models fine-tuned with C-RLFT, an offline reinforcement learning strategy that learns f… | 20 | 5485 | maintenance |
| mattmireles/gemma-tuner-multimodal A Python tool for LoRA fine-tuning of Gemma 4 and 3n models on text, images, and audio using Apple Silicon's Metal Performance Shaders. It … | 61 | 1504 | active |
| open-mmlab/mmengine MMEngine is the foundational training engine library for OpenMMLab projects, providing a unified training loop, config system, registry, ho… | 68 | 1496 | active |
| AI4Finance-Foundation/FinNLP FinNLP is a Python library for collecting internet-scale financial data from sources like Finnhub, Yahoo Finance, Reuters, and Sina Finance… | 31 | 1491 | active |
| alxndrTL/mamba.py A simple, readable pure-PyTorch (plus MLX) implementation of the Mamba state-space model architecture with a parallel scan for efficient tr… | 50 | 1480 | active |
| salesforce/CodeTF CodeTF is a Python transformer library for code large language models, providing unified interfaces for training, fine-tuning, and inferenc… | 10 | 1478 | active |
| Om-Alve/smolGPT A minimal pure-PyTorch implementation for training small GPT-style LLMs from scratch, featuring flash attention, RMSNorm, SwiGLU, RoPE, and… | 21 | 1474 | active |
| facebookresearch/MobileLLM Meta's training code for MobileLLM, a family of sub-billion parameter language models optimized for on-device use, published at ICML 2024. … | 56 | 1465 | active |
| tianweiy/DMD2 DMD2 is the official PyTorch implementation of Improved Distribution Matching Distillation, a NeurIPS 2024 method that distills diffusion m… | 25 | 1464 | active |
| Danau5tin/multi-agent-coding-system A Python multi-agent AI coding system where an orchestrator agent dispatches explorer and coder subagents with shared knowledge artifacts, … | 28 | 1463 | active |
| tensorflow/tpu A collection of reference models and tools for training machine learning models on Google Cloud TPUs, maintained as a public mirror by the … | 68 | 5278 | maintenance |
| ByteDance-Seed/m3-agent M3-Agent is a multimodal agent framework from ByteDance Seed that processes real-time visual and auditory inputs to build entity-centric lo… | 45 | 1459 | active |
| microsoft/KBLaM Official implementation of KBLaM, a method for augmenting pre-trained LLMs with external knowledge by encoding a knowledge base into contin… | 66 | 1452 | active |
| deepseek-ai/EPLB EPLB is DeepSeek's open-source Expert Parallelism Load Balancer for Mixture-of-Experts models. It computes balanced expert replication and … | 23 | 1443 | active |
| SakanaAI/evolutionary-model-merge Official repository for SakanaAI's Evolutionary Model Merge research, providing code and resources to reproduce paper evaluations of models… | 16 | 1442 | active |
| Agent-RL/ReCall ReCall is a framework that trains LLMs to reason with arbitrary tool calls via reinforcement learning, without supervised tool-use trajecto… | 26 | 1441 | active |
| HongyuanLuke/frequencylaw Official code repository for the paper 'Textual Frequency Law on Large Language Models', implementing TFL, TFD, and CTFT methods for studyi… | 46 | 1428 | active |
| microsoft/rStar Microsoft's research repository for rStar2-Agent, a 14B math reasoning model trained with agentic reinforcement learning that autonomously … | 38 | 1427 | active |
| ARahim3/mlx-tune A Python library for fine-tuning LLMs, vision-language, audio (TTS/STT), embedding, OCR, and JEPA models natively on Apple Silicon Macs usi… | 70 | 1416 | active |
| lucidrains/self-rewarding-lm-pytorch A PyTorch library implementing the Self-Rewarding Language Model training framework from MetaAI, along with the SPIN training method. It pr… | 17 | 1412 | active |
| lucidrains/transfusion-pytorch A PyTorch implementation of Transfusion, MetaAI's approach to predicting the next token and diffusing images with a single multi-modal mode… | 70 | 1408 | active |
| SpikingBrain SpikingBrain-7B is a brain-inspired large language model that combines hybrid efficient attention, MoE modules, and spike encoding, with a … | 51 | 1408 | active |
| open-gigaai/giga-world-policy GigaWorld-Policy is a World Action Model (WAM) for robot policy learning that jointly models actions and future visual observations during … | 53 | 1406 | active |
| RahulSChand/gpu_poor A web-based calculator that estimates GPU memory requirements and inference/finetuning throughput (token/s) for any LLM. It supports quanti… | 19 | 1401 | active |
| McGill-NLP/webllama WebLlama is a framework for building Llama-3-powered agents that browse the web by following natural language instructions and dialogue. It… | 56 | 1400 | active |
| zjunlp/KnowLM KnowLM is an open-source framework for building knowledgeable large language models, covering data processing, pre-training, fine-tuning, k… | 31 | 1386 | active |
| keyu-tian/SparK SparK is the official PyTorch implementation of an ICLR 2023 Spotlight paper that applies BERT/MAE-style masked image modeling to convoluti… | 22 | 1382 | stable |
| AlmondGod/tinyworlds A minimal Python implementation of DeepMind's Genie autoregressive world model, including a video tokenizer, action tokenizer, and dynamics… | 51 | 1380 | active |
| macanv/BERT-BiLSTM-CRF-NER A TensorFlow implementation of named entity recognition that fine-tunes Google BERT with a BiLSTM-CRF model, primarily targeting Chinese te… | 32 | 4907 | maintenance |
| OpenRaiser/NanoResearch NanoResearch is an autonomous AI research assistant that runs an end-to-end pipeline from a research idea to a complete LaTeX paper. It exe… | 57 | 1369 | active |
| bytedance/flux Flux is a GPU kernel library from ByteDance that overlaps computation with communication for tensor and expert parallelism in dense and MoE… | 30 | 1369 | active |
| FreedomIntelligence/HuatuoGPT-o1 HuatuoGPT-o1 is a medical large language model project for complex medical reasoning, with open-sourced models (7B-72B), fine-tuning data, … | 22 | 1363 | active |
| huggingface/finetrainers finetrainers is a Hugging Face library for scalable, memory-optimized training (fine-tuning) of diffusion models, including LoRA training o… | 68 | 1362 | active |
| llm-attacks/llm-attacks Official research code for 'Universal and Transferable Adversarial Attacks on Aligned Language Models', implementing the GCG algorithm for … | 29 | 4839 | maintenance |
| MoonshotAI/Kimi-Dev Kimi-Dev-72B is an open-source coding LLM from Moonshot AI specialized in software engineering tasks like issue resolution and bug fixing, … | 35 | 1355 | active |
| MegEngine/MegEngine MegEngine is a fast, scalable deep learning framework with automatic differentiation, developed in C++ with Python bindings. It unifies tra… | 23 | 4810 | maintenance |
| ace-agent/ace ACE is a Python framework for Agentic Context Engineering that lets language models self-improve by treating contexts as evolving playbooks… | 58 | 1344 | active |
| alibaba/graph-learn Graph-Learn (formerly AliGraph) is a distributed framework for developing and applying large-scale graph neural networks, with a training l… | 33 | 1340 | active |
| BytedTsinghua-SIA/CUDA-Agent CUDA-Agent is a large-scale agentic reinforcement learning system from ByteDance Seed and Tsinghua that trains LLMs to generate high-perfor… | 53 | 1338 | active |
| CarperAI/trlx trlX is a distributed training framework for fine-tuning large language models with reinforcement learning from human feedback (RLHF), supp… | 23 | 4756 | maintenance |
| LLaVA-VL/LLaVA-NeXT LLaVA-NeXT is a collection of open large multimodal models (LLaVA-NeXT, LLaVA-Video, LLaVA-OneVision, LLaVA-Critic-R1) that combine vision … | 61 | 4729 | maintenance |
| Vahe1994/AQLM Official PyTorch implementation of AQLM, an extreme LLM compression method via additive quantization, extended with PV-Tuning for finetunin… | 54 | 1329 | active |
| Tongjilibo/bert4torch bert4torch is a PyTorch library providing an elegant reimplementation of transformer models (BERT, RoBERTa, T5, GPT, ChatGLM, LLaMA, etc.) … | 76 | 1326 | active |
| Duxiaoman-DI/XuanYuan XuanYuan is a family of open-source Chinese financial-domain large language models from Duxiaoman, including base, chat, and quantized vari… | 30 | 1324 | active |
| ChenmienTan/RL2 RL2 (Ray Less Reinforcement Learning) is a concise Python library for post-training large language models with reinforcement learning, SFT,… | 54 | 1316 | active |
| Alibaba-NLP/ZeroSearch ZeroSearch is a reinforcement learning framework from Alibaba's Tongyi Lab that trains LLMs to use search by simulating search engine resul… | 33 | 1313 | active |
| deepseek-ai/DeepSeek-Prover-V2 DeepSeek-Prover-V2 is an open-source large language model for formal theorem proving in Lean 4, trained via reinforcement learning with sub… | 31 | 1311 | active |
| greyhaven-ai/autocontext Autocontext is a recursive self-improving harness that runs AI agents against evaluations, retains useful lessons, and produces traces, rep… | 60 | 1309 | active |
| aws-samples/claude-prompt-generator A Python web application that generates, translates, and iteratively evaluates prompts for Anthropic Claude 3 via AWS Bedrock, including co… | 10 | 1309 | active |
| SakanaAI/text-to-lora Text-to-LoRA (T2L) is a hypernetwork that generates LoRA adapters for large language models in a single forward pass, using only a natural … | 28 | 1307 | active |
| derrian-distro/LoRA_Easy_Training_Scripts A PySide6 desktop GUI that wraps Kohya's sd-scripts to simplify training LoRA, LoCon, and other LoRA-type models for Stable Diffusion. It s… | 30 | 1304 | active |
| yizhongw/self-instruct Self-Instruct is a framework and research codebase for aligning pretrained language models with instructions using data generated by the mo… | 32 | 4612 | maintenance |
| buoyancy99/diffusion-forcing Official research code for the NeurIPS paper 'Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion', implementing a metho… | 62 | 1299 | active |
| X-Square-Robot/wall-x Wall-X is the open-source training and inference stack for X Square Robot's WALL series of embodied foundation models (VLAs) for general-pu… | 61 | 1283 | active |
| DreamLM/Dream Dream 7B is an open diffusion large language model (dLLM) with base and instruct checkpoints, plus inference and training code built on Hug… | 40 | 1279 | active |
| DeepEyes DeepEyes is a research project that trains multimodal vision-language models to 'think with images' using end-to-end reinforcement learning… | 41 | 1278 | active |
| Tencent-Hunyuan/SRPO SRPO is Tencent Hunyuan's research code for fine-tuning diffusion image generation models (e.g., FLUX.1.dev) by aligning the full diffusion… | 51 | 1276 | active |
| ModelCloud/GPTQModel GPTQModel is a production-ready Python toolkit for quantizing (compressing) large language models using GPTQ, AWQ, and related methods, wit… | 92 | 1274 | active |
| aiming-lab/Agent0 Agent0 Series is a research framework for training self-evolving LLM agents from zero external data via tool-integrated reasoning and co-ev… | 55 | 1271 | active |
| lucidrains/flamingo-pytorch A PyTorch implementation of DeepMind's Flamingo visual language model architecture, providing the Perceiver Resampler and Gated Cross-Atten… | 23 | 1267 | active |
| ElectricAlexis/NotaGen NotaGen is a symbolic music generation model that produces high-quality classical sheet music using LLM-style training paradigms: pre-train… | 29 | 1236 | active |
| InternScience/GraphGen GraphGen is a Python framework for knowledge-graph-guided synthetic data generation for LLM training. It builds fine-grained knowledge grap… | 53 | 1232 | active |
| OpenTSLM/OpenTSLM OpenTSLM is a family of Time-Series Language Models that integrate time series as a native modality into pretrained LLMs (Llama, Gemma), en… | 58 | 1230 | active |
| sentient-agi/EvoSkill EvoSkill is an open-source Python framework that automatically discovers and synthesizes reusable skills from failed agent trajectories to … | 72 | 1229 | active |
| alibaba/x-deeplearning X-DeepLearning (XDL) is an industrial deep learning framework from Alibaba optimized for high-dimension sparse data scenarios such as adver… | 23 | 4301 | maintenance |
| run-house/kubetorch Kubetorch is a Python library that lets you distribute and run ML workloads (training, inference, data processing) on Kubernetes directly f… | 76 | 1225 | active |
| sgl-project/SpecForge SpecForge is a Python framework from the SGLang team for training speculative decoding models such as EAGLE/EAGLE3 draft heads. Trained mod… | 64 | 1225 | active |
| Bolin97/GongBU GongBU is a self-hosted, no-code web platform for fine-tuning, evaluating, and deploying large language models, built on Transformers and P… | 49 | 1225 | active |
| EvolvingLMMs-Lab/LLaVA-OneVision-2 A fully open framework for training multimodal large language models, releasing models, datasets, and training recipes for the LLaVA-OneVis… | 70 | 1220 | active |
| yeyupiaoling/Whisper-Finetune A toolkit for fine-tuning OpenAI's Whisper speech recognition models using LoRA, supporting training with or without timestamps and even wi… | 63 | 1219 | active |
| Cerebras/modelzoo Cerebras Model Zoo is a collection of reference deep learning model implementations (Llama, Mixtral, DINOv2, Llava, etc.) with configs and … | 74 | 1202 | active |
| autonomousvision/stylegan-t Official training code for StyleGAN-T, an ICML 2023 paper on fast large-scale text-to-image synthesis using GANs. It provides dataset prepa… | 32 | 1197 | active |
| DeepRec-AI/DeepRec DeepRec is a high-performance deep learning framework for recommendation models, built on TensorFlow 1.15 with Intel and NVIDIA TensorFlow … | 23 | 1197 | active |
| facebookresearch/esm Meta FAIR's Evolutionary Scale Modeling (ESM) library providing Transformer protein language models with pretrained weights, including ESM-… | 10 | 4170 | maintenance |
| Tencent-Hunyuan/MixGRPO MixGRPO is a research framework from Tencent Hunyuan implementing a mixed ODE-SDE GRPO algorithm for efficient reinforcement learning fine-… | 56 | 1181 | active |
| MediaBrain-SJTU/MING MING (明医) is a Chinese medical consultation large language model fine-tuned on medical instruction data, with variants built on bloomz-7b a… | 37 | 1181 | active |
| Facico/Chinese-Vicuna Chinese-Vicuna is a low-resource LLaMA+LoRA solution for building Chinese instruction-following language models, structured after Alpaca. I… | 35 | 4106 | maintenance |
| princeton-nlp/MeZO MeZO is a memory-efficient zeroth-order optimizer that fine-tunes language models using only forward passes, with the same memory footprint… | 30 | 1174 | stable |
| baichuan-inc/Baichuan2 Baichuan 2 is a family of open large language models (7B and 13B, Base and Chat variants with 4-bit quantized versions) trained by Baichuan… | 28 | 4082 | maintenance |
| deepmodeling/Uni-Mol Uni-Mol is a collection of 3D molecular representation learning frameworks and pretrained models for tasks like molecule property predictio… | 30 | 1167 | active |
| facebookresearch/fairseq2 fairseq2 is a PyTorch-based sequence modeling toolkit from Meta FAIR for training custom models for content generation tasks such as langua… | 86 | 1152 | active |
| open-gigaai/giga-world-1 GigaWorld-1 is an open-source framework providing training, inference, data processing, checkpoint conversion, and LoRA merge workflows for… | 52 | 1152 | active |
| brightmart/albert_zh A repository providing pre-trained ALBERT models for Chinese language, implemented in TensorFlow with PyTorch and Keras conversions. It inc… | 32 | 3984 | maintenance |
| Being-H Being-H is a family of human-centric embodied foundation models, including VLA models (Being-H0.5, Being-H0) and latent world-action models… | 62 | 1148 | active |
| amazon-science/mm-cot Official PyTorch implementation of the paper 'Multimodal Chain-of-Thought Reasoning in Language Models', which adds vision features to a tw… | 31 | 3978 | maintenance |
| MoonshotAI/MoonEP MoonEP is an Expert Parallelism communication library for Mixture-of-Experts training that keeps token loads perfectly balanced across rank… | 56 | 1146 | active |
| THUMNLab/AutoGL AutoGL is an autoML framework and toolkit for machine learning on graphs, built on PyTorch with PyTorch Geometric and DGL backends. It prov… | 43 | 1142 | active |
| LLM-Pruner LLM-Pruner is a PyTorch library implementing structural pruning of large language models based on gradient information, as published at Neu… | 30 | 1142 | active |
| flagos-ai/FlagGems FlagGems is a high-performance operator library for large language models written in the Triton language, providing backend-neutral GPU ker… | 88 | 1137 | active |
| OpenDriveLab/UniVLA UniVLA is an open-source framework for training cross-embodiment vision-language-action (VLA) robot policies using task-centric latent acti… | 40 | 1134 | active |
| NovaSearch-Team/RAG-Retrieval A Python library and toolkit for unified fine-tuning, inference, and distillation of RAG retrieval models, including embedding models, ColB… | 58 | 1131 | active |