Open-Source Training and RLHF Frameworks

XTuner is an LLM training engine for researchers working with very large mixture-of-experts (MoE) models. Its focus is reducing memory pressure and improving distributed training efficiency when model experts receive uneven workloads.

OLMo Core is a Python library for researchers and developers building and training OLMo language models. Its PyTorch components and official training scripts suit teams working with model training code on their own hardware or servers.

Flower is a federated AI framework for developers and researchers who want to train models across separate organizations or devices. Participants retain control of their data rather than gathering everyone's data in one place.

Axolotl is a free, open-source LLM fine-tuning framework for developers and researchers adapting models to their own data. It supports training on local hardware or cloud infrastructure you control.

PaddleSeg is an image segmentation toolkit for developers building computer vision applications on their own hardware. It covers annotation, training and deployment for tasks such as medical image analysis, industrial inspection and portrait effects.

CatBoost is an open-source machine learning library for developers and data scientists building prediction models with numerical and categorical data. Its built-in handling of categories reduces the work needed to turn non-numeric inputs into usable features. It runs computations on your own CPU or GPU and uses the Apache 2.0 license.

OpenRLHF is a self-hosted Python framework for researchers and teams training language models with human feedback or custom rewards. It supports fine-tuning and reinforcement learning on your own GPU hardware, including training agents that interact with an environment.

Ludwig is a framework for developers and researchers training custom AI models without writing their own training loops. It handles preprocessing, training and evaluation from a model definition, covering LLMs as well as tabular and multimodal tasks.

SpecForge is an open-source framework for teams training draft models to speed up LLM inference in SGLang. It suits developers who want to train on their own hardware and use the resulting drafts directly with SGLang serving.

Accelerate is a Python library for developers and researchers who write their own PyTorch training loops. It handles hardware and distributed execution so they can keep control of how their models train.

NVIDIA FLARE is a Python SDK for researchers and data scientists who want to train models across participating sites while keeping training data local. Platform developers can also use it to build systems for collaboration between organizations. It supports on-premise and cloud deployments and is open source under the Apache 2.0 license.

GPT-NeoX is EleutherAI's framework for researchers and teams training language models with billions of parameters on their own GPU infrastructure. It focuses on training from scratch, with support for fine-tuning existing checkpoints.

Colossal-AI is a Python framework for developers and researchers training or serving large AI models on their own GPUs. It provides tools to spread workloads across GPUs and work within available memory.

AgileRL is an open-source Python framework for teams training AI agents on their own tasks, including LLM agents and agents for robotics, finance or logistics. Its evolutionary hyperparameter tuning adjusts training settings during a run, reducing the need for separate tuning trials. The framework uses the Apache 2.0 license.

LightGBM is a decision-tree gradient boosting framework for classification, regression and ranking. It trains models on your own machine or across multiple machines, with GPU support for compatible configurations. The code is open source under the MIT license.

NVIDIA PhysicsNeMo is an open-source Python framework for engineers and researchers building AI models of physical systems. It combines physics constraints with simulation and observed data to train surrogate models that can predict behavior in near real time. The framework uses PyTorch and carries the Apache 2.0 license.

tiny-cuda-nn is a neural network framework for developers and researchers building small models, NeRF renderers and neural graphics applications. It handles training and inference on local NVIDIA GPUs.

PEFT is a library for developers adapting pretrained AI models on their own hardware. It trains a small portion of a model's parameters to reduce the compute and storage costs of full fine-tuning.

FederatedScope is an open-source federated learning platform for researchers and teams building training systems across separate data holders. It can simulate clients and servers on one device or run them as separate distributed participants. Each participant loads its own data, while models can pass between participants through messages.

DeepReg is an open-source Python toolkit for researchers and educators studying medical image registration with deep learning. It brings together algorithms for aligning medical images, with an emphasis on clinical applications and methods that use limited supervision.

PaddleDetection is a computer vision toolkit for developers building image and video analysis systems. It covers training, adapting and deploying models for uses such as inspection and traffic analysis.

DeepCell is an open-source Python library for researchers analyzing individual cells in biological images. It runs locally and uses TensorFlow to identify whole cells or nuclei and track cells through time-lapse movies. Its Apache 2.0 license allows users to adapt the library for their own research.

Substra is open-source federated learning software for teams that need to train and validate machine learning models across distributed datasets. It supports production use across a network, as well as simulations on a single machine for testing and debugging federated learning code.

Molmo is Ai2's family of vision-language models for working with images and text on your own hardware. It's aimed at developers and researchers who want to train models or evaluate them on visual tasks.

SimpleTuner is a diffusion model training toolkit for creators and researchers adapting image, video or audio models to their own datasets. Its web dashboard also supports teams sharing GPU workers.

Keras is an open-source Python deep learning framework for developers and researchers who want to build and train models without tying their code to one backend. It runs locally on Linux and macOS, with WSL2 recommended for Windows. Its Apache 2.0 license permits commercial use and modification.

Stanza is a Python NLP library for developers and researchers who need to analyze text in multiple human languages on their own hardware. It turns raw text into linguistic annotations, such as word forms, grammatical relationships, and named entities. It's open source under Apache 2.0.

AI Toolkit (ostris) is a training suite for people fine-tuning image and video generation models on their own hardware or rented servers. It provides a browser interface and command-line access to training jobs.

diffusion-pipe trains image and video diffusion models on your own GPU hardware. It's for people fine-tuning models that are too large to fit on a single GPU or who want to distribute training across several GPUs.

verl is a Python library for researchers and engineers training large language models on their own hardware. It implements HybridFlow for reinforcement learning after initial training, with support for supervised fine-tuning too.

scikit-learn is a Python machine learning library for people building predictive models and analyzing data. It combines algorithms with data preparation and evaluation tools so you can compare approaches within one library.

Optuna is a Python framework for machine learning developers and researchers who want to automate the search for model settings. Its search spaces can change with the choices made during a trial, so it can compare model architectures or classifiers as well as tune individual parameters. It's open source under the MIT license and can run on your own hardware as part of a Python workload.

Dask-ML is a Python machine learning library for people whose datasets exceed RAM or whose training and evaluation jobs take too long. It uses Dask to distribute work across a cluster of machines, while keeping interfaces familiar to users of Scikit-Learn, NumPy, and pandas. It's open source under the BSD-3-Clause license.

SLEAP helps behavioral researchers measure animal movement in recorded video. It trains models to estimate poses for single animals or groups, with human review during dataset creation and prediction checking.