stability-ai/stablelm

★ 15,709Jupyter NotebookAudience · researcherComplexity · 4/5LicenseSetup · hard

Mindmap

mindmap
  root((StableLM))
    What it does
      Open-source language models
      Text generation
      Instruction following
    Model variants
      3B parameter model
      7B parameter model
      StableVicuna
    Use cases
      Local LLM inference
      Fine-tuning
      Research experiments
    Tech stack
      PyTorch
      Jupyter Notebooks
      Hugging Face

mindmap root((StableLM)) What it does Open-source language models Text generation Instruction following Model variants 3B parameter model 7B parameter model StableVicuna Use cases Local LLM inference Fine-tuning Research experiments Tech stack PyTorch Jupyter Notebooks Hugging Face

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Run an open-source language model locally for text generation without relying on a paid API.

USE CASE 2

Fine-tune StableLM on a custom dataset to create a domain-specific AI assistant.

USE CASE 3

Study and experiment with open-source LLM architectures and training approaches using the included Jupyter Notebooks.

USE CASE 4

Use the efficient 3B model on limited hardware where larger models would be too slow or expensive.

Tech stack

PythonPyTorchJupyter Notebook

Getting it running

Difficulty · hard Time to first run · 1h+

Requires a GPU with sufficient VRAM and downloading large model checkpoint files from Hugging Face.

Released under CC BY-SA-4.0, free to use and adapt with attribution, but any modified versions must be shared under the same share-alike license.

In plain English

StableLM is a series of open-source language models (AI models that can understand and generate text) developed by Stability AI. This repository tracks the ongoing release of different model checkpoints, snapshots of a trained model at various sizes and training stages. A "language model" is the kind of AI that powers chatbots and text generation tools. These models are trained on large amounts of text data so they can answer questions, write, summarize, and more. "Parameters" roughly indicate model size and capability, more parameters generally means a more capable (but slower and more resource-intensive) model. The repository includes several model variants: the StableLM-3B-4E1T (a 3-billion-parameter model trained on 4 trillion tokens, individual pieces of text, across multiple passes), older StableLM-Alpha models at 3B and 7B parameter sizes, and StableVicuna, a version further trained to follow human instructions. The 3B model is notably efficient, a smaller model trained extensively to match the quality of much larger models, making it more practical to run on limited hardware. You would use this repository if you are a researcher or developer who wants to run, fine-tune (customize for a specific task), or study open-source language models as an alternative to commercial AI services. The base models are released under a Creative Commons license (CC BY-SA-4.0), meaning you can use and adapt them with attribution. Jupyter Notebooks are included for experimentation. The full README is longer than what was provided.

Copy-paste prompts

Prompt 1

How do I load the StableLM-3B-4E1T model with Hugging Face Transformers and generate text from a prompt in Python?

Prompt 2

Show me how to fine-tune StableLM on a small custom instruction dataset using LoRA to create a task-specific assistant.

Prompt 3

How do I run StableVicuna locally and have a multi-turn conversation with it from a Python script?

Prompt 4

What are the minimum GPU VRAM requirements to run the 7B StableLM model, and how do I quantize it to reduce memory usage?

Open on GitHub → Explain another repo

← stability-ai on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.