Dev Tools
rocketride-server
A high-performance AI pipeline engine - build LLM workflows as portable .pipe files
★ 17.8k starsMIT
View guide →20+ high-performance LLMs implemented from scratch, with proven recipes to pretrain, finetune and deploy.
litgpt is an open-source dev tools project by Lightning-AI (Lightning-AI/litgpt on GitHub). It has 13,695 GitHub stars and 1,516 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 7 October 2026.
Pick from 20+ supported LLMs and decide whether you are chatting, fine-tuning, pretraining, evaluating or serving.
Install the package with the extra dependencies bundle from PyPI, or clone the repo for a source install.
Use the CLI or the Python API - weights download automatically for supported models (some gated models need a Hugging Face token).
Fine-tune on your JSON dataset with a proven recipe, chat with the result to test it, then serve it as an API.
Every LLM is implemented from scratch with no abstractions, making the code easy to read, debug and modify.
Llama 2/3, Code Llama, Gemma 2/3, Phi, Qwen2.5/Qwen3, Mistral/Mixtral, Falcon, Pythia, OLMo, DeepSeek R1 distills and more, listed via litgpt download list.
Pretrain, continued pretrain, finetune, evaluate (for example MMLU and TruthfulQA), chat and serve, all through one litgpt command interface.
LoRA, QLoRA, Adapter and Adapter v2 recipes, plus YAML configuration files for reproducible, out-of-the-box performance.
Flash Attention v2, FSDP multi-GPU support, optional CPU offloading, TPU/XLA support, and FP16/BF16 plus 4-bit and 8-bit quantization to cut memory.
litgpt serve turns a pretrained or fine-tuned checkpoint into a web server you can query over HTTP.
Clone LitGPT for a source install; PyPI users can skip the clone.
git clone https://github.com/Lightning-AI/litgpt.git
cd litgptInstall the package with its extra dependencies from PyPI.
pip install 'litgpt[extra]'Download and chat with a model in one command - weights are fetched automatically.
litgpt chat microsoft/phi-2LitGPT is developed by Lightning AI, the company behind PyTorch Lightning, with open-source contributors.
Yes. It is Apache-2.0 licensed, which the project highlights as allowing unlimited enterprise use. Model weights you download carry their own licenses.
No. Everything runs on your own GPUs; some gated Hugging Face models require a free access token to download.
Transformers is a broad modeling library and ms-swift a ModelScope training framework; LitGPT focuses on a curated set of from-scratch, hackable implementations with proven end-to-end recipes in one CLI.
Run pip install 'litgpt[extra]', then litgpt chat microsoft/phi-2 to talk to a model in minutes.
A high-performance AI pipeline engine - build LLM workflows as portable .pipe files
A high-performance coding agent harness in Rust - fast TUI, swarms and every provider
Fine-tune and deploy 600+ LLMs and 300+ multimodal models with PEFT or full-parameter training.
NVIDIA NeMo Data Designer: build validated, multimodal synthetic datasets from scratch or from seed data.
Anthropic's official collection of notebooks and recipes for using Claude effectively
Google's Python library for extracting structured data from unstructured text with LLMs, with grounding
The most-starred other Dev Tools projects in the LibreAI catalogue — the realistic shortlist if litgpt is not the right fit.
The open source coding agent
The model-definition framework behind modern AI — thousands of pretrained transformers for text, vision, and audio.
Anthropic's agentic coding tool that lives in your terminal and works across your whole codebase
Cross-platform desktop all-in-one manager for Claude Code, Codex, OpenCode and other coding agents
OpenAI's lightweight coding agent that runs locally in your terminal
Free and open — grab the source, star the repo, and start creating.