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 |
|---|---|---|---|
| AutoArk/open-audio-opd An industrial training stack for online policy distillation (OPD) of audio models, distilling compact ASR (and planned TTS) student models … | 49 | 1128 | active |
| lasgroup/SDPO SDPO (Self-Distilled Policy Optimization) is a research library implementing a reinforcement learning framework for post-training large lan… | 53 | 1126 | active |
| RUC-NLPIR/ARPO ARPO (Agentic Reinforced Policy Optimization) is a reinforcement learning algorithm and training framework for LLM agents, published at ICL… | 52 | 1126 | active |
| PRIME-RL/TTRL TTRL is an open-source implementation of Test-Time Reinforcement Learning, a method for training LLMs with RL on unlabeled data using major… | 49 | 1124 | active |
| datadreamer-dev/DataDreamer DataDreamer is an open-source Python library for prompting LLMs, generating synthetic datasets, and training or aligning models in reproduc… | 32 | 1124 | active |
| THUDM/SwissArmyTransformer SwissArmyTransformer (sat) is a PyTorch library for developing custom Transformer model variants where models like BERT, GPT, T5, GLM, and … | 23 | 1122 | active |
| FlagAI-Open/FlagAI FlagAI is a Python toolkit for training, fine-tuning, and deploying large-scale AI models across NLP, CV, and vision-language tasks. It int… | 61 | 3870 | maintenance |
| HITsz-TMG/Uni-MoE Uni-MoE is a family of open-source Mixture-of-Experts (MoE) based omnimodal large language models that understand and generate across text,… | 65 | 1118 | active |
| yahoo/TensorFlowOnSpark TensorFlowOnSpark is a Python library that lets existing TensorFlow programs run distributed training and inference on Apache Spark and Had… | 23 | 3847 | maintenance |
| BytedTsinghua-SIA/MemAgent MemAgent is a reinforcement-learning framework for training LLM agents that process arbitrarily long contexts via a memory mechanism within… | 52 | 1111 | active |
| yandex/YaLM-100B YaLM-100B is a GPT-like pretrained language model with 100 billion parameters, trained by Yandex on English and Russian text using DeepSpee… | 32 | 3756 | maintenance |
| mymusise/ChatGLM-Tuning A Python toolkit for fine-tuning the ChatGLM-6B large language model using LoRA (Low-Rank Adaptation) with the Alpaca dataset. It provides … | 10 | 3738 | maintenance |
| kimiyoung/transformer-xl Official implementation of Transformer-XL, an attention-based language model architecture that extends context beyond a fixed length via se… | 32 | 3717 | maintenance |
| bytedance/byteps BytePS is a high-performance parameter server framework for distributed deep neural network training, supporting TensorFlow, Keras, PyTorch… | 10 | 3717 | maintenance |
| GAIR-NLP/LIMO LIMO is a research project and training framework demonstrating that large language models can achieve strong mathematical reasoning with o… | 33 | 1083 | active |
| arcee-ai/DistillKit DistillKit is an open-source Python toolkit for knowledge distillation of large language models, supporting both online and offline distill… | 57 | 1076 | active |
| THUDM/GLM GLM is a general language model pretrained with an autoregressive blank-filling objective, released with pretrained checkpoints and fine-tu… | 32 | 3680 | maintenance |
| open-gigaai/giga-train GigaTrain is an efficient and scalable Python training framework for large AI models, supporting distributed multi-GPU/multi-node execution… | 59 | 1072 | active |
| NVlabs/DiffusionNFT DiffusionNFT is a research library implementing an online reinforcement learning paradigm for diffusion models that optimizes policy direct… | 44 | 1070 | active |
| X-LANCE/SLAM-LLM SLAM-LLM is a deep learning toolkit for training custom multimodal large language models focused on speech, language, audio, and music proc… | 52 | 1066 | active |
| zhuzilin/ring-flash-attention A Python library implementing RingAttention on top of FlashAttention for distributed long-context transformer training. It provides varlen … | 38 | 1061 | active |
| open-gigaai/giga-models GigaModels is an open-source Python framework providing pipelines for training, inference, deployment, and compression of multi-modal, gene… | 59 | 1060 | active |
| NousResearch/DisTrO DisTrO is a family of low-latency distributed optimizers that reduce inter-GPU communication requirements by three to four orders of magnit… | 41 | 1059 | active |
| TIGER-AI-Lab/verl-tool VerlTool is a unified, extensible framework built on verl for training LLM agents with tool use via reinforcement learning. It decouples ac… | 61 | 1045 | active |
| NVIDIA-NeMo/Skills Nemo-Skills is a collection of Python pipelines for improving the skills of large language models, covering synthetic data generation, mode… | 69 | 1042 | active |
| kuleshov-group/bd3lms BD3-LMs is a research implementation of Block Discrete Denoising Diffusion Language Models that interpolate between autoregressive and diff… | 31 | 1037 | active |
| yongliang-wu/DFT DFT (Dynamic Fine-Tuning) is the official implementation of an ICLR 2026 paper that improves Supervised Fine-Tuning of LLMs by dynamically … | 58 | 1035 | active |
| kaito-project/kaito KAITO is a Kubernetes operator suite that automates LLM inference, fine-tuning, and RAG engine deployment using simplified CRD APIs. It aut… | 93 | 1028 | active |
| fla-org/native-sparse-attention Efficient Triton kernel implementations of Native Sparse Attention (NSA), a hardware-aligned, natively trainable sparse attention mechanism… | 47 | 1023 | active |
| microsoft/Tutel Tutel is Microsoft's optimized Mixture-of-Experts (MoE) library for efficient training and inference of large language models, featuring dy… | 77 | 1019 | active |
| TRI-ML/prismatic-vlms Prismatic VLMs is a PyTorch-based codebase for training visually-conditioned language models (VLMs) with flexible vision backbones like CLI… | 26 | 1019 | active |
| deepseek-ai/DeepSeek-Math DeepSeekMath is a 7B open language model specialized in mathematical reasoning, initialized from DeepSeek-Coder and trained on 500B math-re… | 26 | 3457 | maintenance |
| tensorflow/adanet AdaNet is a lightweight TensorFlow-based AutoML framework that automatically learns high-quality neural network architectures and ensembles… | 10 | 3451 | maintenance |
| EvolvingLMMs-Lab/Otter Otter is a multi-modal vision-language model built on OpenFlamingo, instruction-tuned on the MIMIC-IT dataset with image and video understa… | 21 | 3445 | maintenance |
| mni-ml/framework A TypeScript machine learning framework with Rust native backends for CPU, CUDA, and WebGPU. It provides autograd, tensor operations, built… | 48 | 1010 | active |
| TinyLLaVA/TinyLLaVA_Factory TinyLLaVA Factory is an open-source modular PyTorch/HuggingFace codebase for training small-scale large multimodal models (LMMs) that combi… | 70 | 1009 | active |
| QwenAudio/Fun-Audio-Chat Fun-Audio-Chat is a Large Audio Language Model (8B) for natural, low-latency voice interactions, using dual-resolution speech representatio… | 60 | 1009 | active |
| argilla-io/distilabel Distilabel is a Python framework for building scalable pipelines that generate synthetic data and AI feedback, based on verified research p… | 74 | 3408 | maintenance |
| minimaxir/gpt-2-simple A Python package that simplifies fine-tuning OpenAI's GPT-2 text-generation model (124M/355M) on custom text and generating text from the r… | 23 | 3405 | maintenance |
| MoonshotAI/checkpoint-engine Checkpoint-engine is a lightweight Python middleware for updating model weights in-place across LLM inference engines, a critical step in r… | 78 | 1005 | active |
| pjlab-sys4nlp/llama-moe LLaMA-MoE is a Python toolkit and model series for building Mixture-of-Experts (MoE) language models from dense LLaMA models via expert con… | 20 | 1005 | active |
| facebookresearch/fairscale FairScale is a PyTorch extension library providing composable modules and APIs for high-performance, large-scale distributed training, incl… | 27 | 3403 | maintenance |
| Liuhong99/Sophia Official PyTorch implementation of Sophia-G, a scalable stochastic second-order optimizer for language model pre-training, from the paper b… | 30 | 1001 | active |
| catalyst-team/catalyst Catalyst is a high-level PyTorch framework for deep learning research and development, focused on reproducibility, rapid experimentation, a… | 60 | 3386 | maintenance |
| WangRongsheng/CareGPT CareGPT is a medical large language model project that aggregates dozens of publicly available medical fine-tuning datasets and open medica… | 10 | 999 | active |
| JIA-Lab-research/MGM Official PyTorch implementation of Mini-Gemini, a multimodal vision-language model framework built on LLaVA that supports dense and MoE LLM… | 26 | 3322 | maintenance |
| XYZ-AI-Lab/axrl AxisRL is an agentic reinforcement learning post-training framework for large language models, built on SGLang for high-throughput rollout … | 53 | 984 | active |
| bytedance/lightseq LightSeq is a high-performance CUDA-based library for training and inference of sequence models like Transformer, BERT, GPT, and BART, with… | 10 | 3292 | maintenance |
| google-research/albert Official TensorFlow implementation and pretrained checkpoints of ALBERT, a lite version of BERT for self-supervised learning of language re… | 10 | 3278 | maintenance |
| JoePenna/Dreambooth-Stable-Diffusion A Jupyter Notebook-based implementation of Dreambooth fine-tuning for Stable Diffusion, adapted from XavierXiao's repo with tweaks for trai… | 32 | 3213 | maintenance |
| huawei-noah/Pretrained-Language-Model A collection of pretrained language models and optimization techniques from Huawei Noah's Ark Lab, including PanGu-α (200B-parameter Chines… | 32 | 3166 | maintenance |
| project-baize/baize-chatbot Baize is an open-source chat model built on LLaMA using LoRA parameter-efficient tuning, trained on 100k self-chat dialogs generated by Cha… | 21 | 3145 | maintenance |
| microsoft/torchscale A PyTorch library from Microsoft implementing foundation Transformer architectures such as DeepNet, Magneto, RetNet, LongNet, BitNet, and X… | 32 | 3141 | maintenance |
| linyiLYi/bilibot A local chatbot fine-tuned from Bilibili user comments, built on Qwen1.5-32B-Chat using Apple's MLX LoRA fine-tuning. It supports text chat… | 25 | 3127 | maintenance |
| dbiir/UER-py UER-py is a PyTorch framework for pre-training transformer language models (BERT, GPT-2, T5, ELMo, etc.) and fine-tuning them on downstream… | 32 | 3114 | maintenance |
| salesforce/CodeT5 Official research release of CodeT5 and CodeT5+ open code large language models from Salesforce Research for code understanding and generat… | 10 | 3093 | maintenance |
| rinongal/textual_inversion Official implementation of the Textual Inversion paper, which learns new word embeddings in a frozen text-to-image (Latent Diffusion) model… | 32 | 3055 | maintenance |
| T5 T5X is a modular, composable framework built on JAX and Flax for high-performance training, evaluation, and inference of sequence models at… | 71 | 3002 | maintenance |
| yangjianxin1/GPT2-chitchat A GPT2-based Chinese chitchat dialogue model project built on HuggingFace transformers, including training, preprocessing, and interactive … | 32 | 2995 | maintenance |
| FreedomIntelligence/LLMZoo LLM Zoo is a project providing data, models, and evaluation benchmarks for large language models, including the multilingual Phoenix and Ch… | 31 | 2936 | maintenance |
| facebookresearch/XLM PyTorch implementation of Cross-lingual Language Model Pretraining (XLM) from Facebook AI Research, covering MLM, CLM, and TLM objectives p… | 10 | 2922 | maintenance |
| Tencent/PocketFlow PocketFlow is an open-source AutoML framework from Tencent AI Lab for automatically compressing and accelerating deep learning models. Deve… | 32 | 2910 | maintenance |
| cybertronai/gradient-checkpointing A Python library that reduces GPU memory usage when training very deep neural networks via gradient checkpointing, trading computation for … | 32 | 2847 | maintenance |
| OpenPipe/OpenPipe OpenPipe is an open-source fine-tuning and model-hosting platform that turns expensive LLM prompts into cheaper fine-tuned models. It offer… | 30 | 2832 | maintenance |
| Alpha-VLLM/LLaMA2-Accessory LLaMA2-Accessory is an open-source Python toolkit for pretraining, finetuning, and deploying large language models and multimodal LLMs, inc… | 29 | 2796 | maintenance |
| PhoebusSi/Alpaca-CoT Alpaca-CoT is an instruction-tuning platform that unifies interfaces for instruction data collection, parameter-efficient fine-tuning metho… | 31 | 2788 | maintenance |
| liucongg/ChatGLM-Finetuning A Python toolkit for fine-tuning ChatGLM-6B, ChatGLM2-6B, and ChatGLM3-6B large language models using Freeze, LoRA, P-Tuning, and full-para… | 21 | 2773 | maintenance |
| kyegomez/OpenMythos OpenMythos is an open-source, theoretical PyTorch implementation of a Recurrent-Depth Transformer (RDT) architecture inspired by speculatio… | 49 | 14910 | experimental |
| JIA-Lab-research/LongLoRA LongLoRA is an efficient fine-tuning approach and codebase that extends the context length of pre-trained LLMs (Llama2 7B/13B/70B) using sh… | 28 | 2687 | maintenance |
| alibaba/pipcook Pipcook is a JavaScript application framework for machine learning and its engineering, aimed at enabling JavaScript and front-end engineer… | 67 | 2596 | maintenance |
| jcjohnson/torch-rnn torch-rnn provides high-performance, reusable RNN and LSTM modules for the Torch7 deep learning framework, used for character-level languag… | 32 | 2562 | maintenance |
| LiyuanLucasLiu/RAdam RAdam is a Python implementation of Rectified Adam, a variant of the Adam optimizer that analytically reduces the large variance of adaptiv… | 32 | 2550 | maintenance |
| automl/Auto-PyTorch Auto-PyTorch is an AutoML framework that jointly performs neural architecture search and hyperparameter optimization for PyTorch models, us… | 23 | 2540 | maintenance |
| young-geng/EasyLM EasyLM is a JAX/Flax-based framework for pre-training, finetuning, evaluating, and serving large language models like LLaMA. It scales trai… | 32 | 2515 | maintenance |
| microsoft/Graphormer Graphormer is a deep learning library from Microsoft providing a graph transformer backbone for molecular modeling tasks. It ships pre-trai… | 58 | 2473 | maintenance |
| microsoft/DialoGPT DialoGPT is a large-scale pretrained dialogue response generation model from Microsoft, based on GPT-2 and trained on 147M multi-turn Reddi… | 10 | 2419 | maintenance |
| OpenBMB/CPM-Bee CPM-Bee is a fully open-source, commercially usable 10-billion-parameter bilingual (Chinese-English) foundation large language model traine… | 68 | 2408 | maintenance |
| allenai/RL4LMs RL4LMs is a modular Python library from AllenAI for fine-tuning language models with reinforcement learning to align with human preferences… | 32 | 2396 | maintenance |
| Texar Texar is a modularized Python toolkit for machine learning, especially natural language processing and text generation, built on TensorFlow… | 61 | 2390 | maintenance |
| TigerResearch/TigerBot TigerBot is a multi-language, multi-task large language model project from TigerResearch, providing pretrained and chat-tuned model weights… | 30 | 2262 | maintenance |
| XLabs AI Flux Toolkit XLabs AI's training scripts for fine-tuning the FLUX.1 diffusion model with LoRA, ControlNet, and IP-Adapter adapters, using DeepSpeed and … | 24 | 2227 | maintenance |
| epfLLM/meditron Meditron is a suite of open-source medical large language models (7B and 70B) adapted from Llama-2 via continued pretraining on a curated m… | 27 | 2211 | maintenance |
| lucidrains/reformer-pytorch A PyTorch implementation of the Reformer, an efficient Transformer architecture using LSH attention, reversible networks, and chunking to h… | 23 | 2195 | maintenance |
| alibaba/EasyNLP EasyNLP is a comprehensive PyTorch-based NLP toolkit from Alibaba that provides training, inference, and deployment for pre-trained languag… | 23 | 2186 | maintenance |
| ai-forever/ru-gpts A repository of Russian GPT-3 language models (ruGPT3XL/Large/Medium/Small and ruGPT2Large) with usage and fine-tuning examples. It provide… | 32 | 2089 | maintenance |
| THUDM/P-tuning-v2 P-tuning v2 is a Python implementation of deep prompt tuning, applying trainable continuous prompts at every transformer layer so prompt tu… | 32 | 2079 | maintenance |
| neulab/prompt2model Prompt2Model is a Python library that takes a natural language task description (like an LLM prompt) and automatically generates a small, s… | 21 | 2023 | maintenance |
| haitongli/knowledge-distillation-pytorch A PyTorch framework for running knowledge distillation experiments, supporting both shallow (teacher-to-small-CNN) and deep distillation on… | 32 | 1999 | maintenance |
| databricks/spark-deep-learning Deep Learning Pipelines for Apache Spark, now reduced to the HorovodRunner component for distributed deep learning training via Horovod on … | 23 | 1987 | maintenance |
| mit-han-lab/once-for-all Once-for-All (OFA) is a PyTorch library implementing the ICLR 2020 Once-for-All network, which trains a single supernet that can be special… | 23 | 1953 | maintenance |
| danieldjohnson/biaxial-rnn-music-composition A Python implementation of a biaxial recurrent neural network (LSTM-based) trained to generate classical music from MIDI data. It includes … | 32 | 1929 | maintenance |
| Palashio/libra Libra is a Python autoML library that lets users build, train, and evaluate machine learning models with one-line natural-language-style qu… | 37 | 1905 | maintenance |
| minimaxir/aitextgen aitextgen is a Python library for training and generating text with GPT-2 and GPT Neo models, built on PyTorch, Hugging Face Transformers, … | 23 | 1836 | maintenance |
| jquesnelle/yarn Reference implementation of YaRN, an efficient method for extending the context window of large language models, published as an ICLR 2024 … | 29 | 1801 | maintenance |
| zai-org/CogView CogView is a 4-billion-parameter pretrained transformer model for text-to-image image generation, released alongside a NeurIPS 2021 paper. … | 32 | 1799 | maintenance |
| cmu-db/noisepage NoisePage is a relational database management system from Carnegie Mellon University designed for autonomous, self-driving operation, with … | 10 | 1764 | maintenance |
| huggingface/transfer-learning-conv-ai A clean, commented PyTorch codebase for training a dialog/chatbot agent by transfer learning from OpenAI GPT/GPT-2 language models. It repr… | 10 | 1757 | maintenance |
| deepset-ai/FARM FARM is a Python framework for fine-tuning and evaluating transformer-based language models for NLP tasks, with a focus on question answeri… | 10 | 1753 | maintenance |
| airaria/TextBrewer TextBrewer is a PyTorch-based knowledge distillation toolkit for NLP models. It provides an easy-to-use framework implementing state-of-the… | 23 | 1709 | maintenance |
| imcaspar/gpt2-ml GPT2-ML is a TensorFlow-based GPT-2 training and inference project supporting multiple languages, with a focus on Chinese. It provides 1.5B… | 23 | 1703 | maintenance |