Tools to Fine-Tune and Train AI Models Locally

Transformers is a Python library for developers and researchers running pretrained AI models or training models on their own hardware. It covers text, images, audio, video and multimodal tasks through shared model definitions.

MLX is a machine learning array framework for researchers and developers building and testing models on their own hardware. It supports model training and inference, with APIs for numerical work and neural networks.

River is a Python library for models that learn from incoming data one observation at a time. It's for developers and data scientists working with event streams, changing data patterns, or applications that need to update models without revisiting past training data. It runs in your own Python environment on Linux, macOS and Windows, and is open source under the BSD-3-Clause license.

text-generation-webui, also called TextGen, runs language models on your own hardware for chat, writing and application integrations. You can use it as a desktop app or a self-hosted browser interface.

Optimum Quanto is an open-source PyTorch library for developers who want to reduce the memory needed to run models on their own hardware. It handles language models and image-generation models, with integrations for Hugging Face Transformers and Diffusers. The project is in maintenance mode, accepting minor fixes and documentation work rather than major features.

DeepSqueak is a MATLAB application for bioacoustics researchers studying animal vocalizations, including ultrasonic calls from mice and rats. It uses neural networks to detect calls in audio recordings and provides a graphical interface for reviewing detections. The code is open source under the BSD-3-Clause license.

FlashAttention is an attention library for developers training or running transformer models on their own GPUs. It computes exact attention while using less memory than standard PyTorch attention, allowing longer sequences.

DeepXDE is an open-source Python library for researchers and engineers who use neural networks to solve scientific problems. It combines physics-informed neural networks (PINNs) with operator learning, so you can work with equations and physical constraints alongside training data. Its license is LGPL-2.1.

ExecuTorch is an on-device AI runtime for developers adding PyTorch models to apps and embedded devices. It handles inference on the user's hardware across Android, iOS, macOS, Linux and Windows, as well as embedded systems and microcontrollers.

Nemotron is NVIDIA's family of AI models for developers building agents that reason, write code and use tools. Teams can evaluate and adapt the models using published weights, training data and training recipes.

CTranslate2 is an inference library for developers who want to run Transformer models on their own hardware or servers. It focuses on reducing execution time and memory use for applications built in C++ or Python.

HarnessX is a Python framework for developers building AI agents whose behavior they can reuse and adapt. It separates model choice from tool use, memory and context handling, so changing an agent's role doesn't require rewriting it.

AutoKeras is a Python AutoML library built on Keras for automated deep learning. It searches for neural-network configurations and trains models from your data, reducing the amount of manual architecture design needed for an experiment.

Kubeflow is a self-hosted AI platform for teams developing and running machine learning workloads on Kubernetes. Its modular projects cover interactive development, training and the work of moving models into production.

tsai is a Python library for researchers and developers building machine learning models for time series and sequences. Built on PyTorch and fastai, it brings model training, prediction and data preparation into one package. It's open source under Apache 2.0.

TPOT is a Python AutoML library for data scientists and researchers who want to automate machine learning pipeline selection on their own hardware. It uses genetic programming to search candidate pipelines, reducing the manual work of comparing models and tuning their parameters. It's free and open source under LGPL-3.0.

TensorRT is NVIDIA's inference toolkit for developers deploying AI applications on NVIDIA GPUs. Its compilers, runtimes and model optimizations target lower response latency and higher throughput on local machines, self-hosted servers and edge devices.

Heretic modifies language models to reduce safety-related refusals without expensive post-training. It's for people who want fewer restrictions in models they run themselves, without manually tuning model internals.

Instant NGP is neural graphics software for researchers and developers working on 3D reconstruction and rendering. It trains representations of scenes and images on a single GPU, then lets you explore the results interactively.

PixArt-Sigma generates images from text prompts on your own machine or a self-hosted server. It's for developers and researchers who want downloadable image models and the code to train their own.

Acme is a Python reinforcement learning library for researchers who need reference agents and reusable components for their own experiments. Its agents provide baselines for comparing algorithm performance, while the underlying building blocks let researchers develop new approaches without starting from scratch.

Coqui TTS is a text-to-speech toolkit for developers and speech researchers who want to generate voices locally or train their own models. The idiap fork builds on coqui-ai/TTS.

TensorFlow Model Optimization Toolkit is an open-source Python library for developers preparing TensorFlow and Keras models for deployment on mobile, IoT and other edge devices, or in the cloud. It helps reduce inference latency and cost, and adapt models to limits on memory, processing power, energy use and storage.

Open-Sora is an AI video generation project for researchers, developers and creators who want to run or adapt a model on their own GPUs. It focuses on animating reference images with text guidance and also accepts text alone.

GluonTS is a Python library for data scientists and researchers who need forecasts that express uncertainty. Its models predict a probability distribution over future values, so you can examine prediction intervals alongside the forecast. You train and use models within your own Python environment.

OpenCLIP is a Python library for developers and researchers building image-text matching and zero-shot classification applications. It supports pretrained CLIP models and training models on your own hardware.

SetFit is a framework for developers and NLP practitioners who need to train text classifiers with a small amount of labeled data. It adapts pretrained Sentence Transformers to a classification task without requiring handcrafted prompts or words that stand in for class labels. Sentiment analysis is one supported use case.

AutoTS is an open-source Python package for people who need forecasts across many time series without choosing a model and preprocessing method for each one. It runs in your own Python environment and uses automated search to compare forecasting approaches for your data. The license is MIT.

VoiceCraft generates speech from text and edits existing recordings using a few seconds of a speaker's voice as reference. It's for researchers and developers who want to work with voice generation on their own hardware without training a separate model for each speaker.

Optimum is a collection of Python packages for developers adapting Hugging Face models to their deployment hardware. It extends Transformers, Diffusers, TIMM and Sentence Transformers with tools for faster inference and training.

RWKV is a language model architecture for developers and researchers building LLM applications or training models. Its recurrent design keeps inference memory constant as text grows, making it relevant when a growing context cache would strain available memory.

Kaito manages LLM inference, fine-tuning and document retrieval services in Kubernetes. It's for teams that want to run models in their own cluster with less manual work sizing GPU capacity and managing deployments.

Gensim is a free, open-source Python library for developers and researchers working with large text collections on their own machines or servers. It trains models that capture topics and semantic relationships, then uses those representations to find related documents. It runs on Linux, Windows and macOS under the GNU Lesser General Public License v2.1 (LGPL-2.1).

4M is a framework for researchers and developers who want to train and run models that combine images, text and other visual information. A single model can perform vision tasks and generate images from mixed or partial inputs.