Axolotl
Axolotl is an open source tool designed to streamline the fine-tuning of various AI models, offering support for multiple configurations and architectures. It enables users to train models, supporting methods like full fine-tuning, LoRA, QLoRA, ReLoRA, and GPTQ. Users can customize configurations using simple YAML files or command-line interface overrides, and load different dataset formats, including custom or pre-tokenized datasets. Axolotl integrates with technologies like xFormers, Flash Attention, Liger kernel, RoPE scaling, and multipacking, and works with single or multiple GPUs via Fully Sharded Data Parallel (FSDP) or DeepSpeed. It can be run locally or on the cloud using Docker and supports logging results and checkpoints to several platforms. It is designed to make fine-tuning AI models friendly, fast, and fun, without sacrificing functionality or scale.
Learn more
Unsloth
Unsloth is an open source platform designed to accelerate and optimize the fine-tuning and training of Large Language Models (LLMs). It enables users to train custom models, such as ChatGPT, in just 24 hours instead of the typical 30 days, achieving speeds up to 30 times faster than Flash Attention 2 (FA2) while using 90% less memory. Unsloth supports both LoRA and QLoRA fine-tuning techniques, allowing for efficient customization of models like Mistral, Gemma, and Llama versions 1, 2, and 3. Unsloth's efficiency stems from manually deriving computationally intensive mathematical steps and handwriting GPU kernels, resulting in significant performance gains without requiring hardware modifications. Unsloth delivers a 10x speed increase on a single GPU and up to 32x on multi-GPU systems compared to FA2, with compatibility across NVIDIA GPUs from Tesla T4 to H100, and portability to AMD and Intel GPUs.
Learn more
Tinker
Tinker is a training API designed for researchers and developers that allows full control over model fine-tuning while abstracting away the infrastructure complexity. It supports primitives and enables users to build custom training loops, supervision logic, and reinforcement learning flows. It currently supports LoRA fine-tuning on open-weight models across both LLama and Qwen families, ranging from small models to large mixture-of-experts architectures. Users write Python code to handle data, loss functions, and algorithmic logic; Tinker handles scheduling, resource allocation, distributed training, and failure recovery behind the scenes. The service lets users download model weights at different checkpoints and doesn’t force them to manage the compute environment. Tinker is delivered as a managed offering; training jobs run on Thinking Machines’ internal GPU infrastructure, freeing users from cluster orchestration.
Learn more
Helix AI
Build and optimize text and image AI for your needs, train, fine-tune, and generate from your data. We use best-in-class open source models for image and language generation and can train them in minutes thanks to LoRA fine-tuning. Click the share button to create a link to your session, or create a bot. Optionally deploy to your own fully private infrastructure. You can start chatting with open source language models and generating images with Stable Diffusion XL by creating a free account right now. Fine-tuning your model on your own text or image data is as simple as drag’n’drop, and takes 3-10 minutes. You can then chat with and generate images from those fine-tuned models straight away, all using a familiar chat interface.
Learn more
