Ollama

The easiest way to run ai models.

Download

• macOS (Apple Silicon)
• macOS (Intel – Coming soon)
• Windows (Coming soon)
• Linux (Coming soon)

Python SDK

pip install ollama

Python SDK quickstart

import ollama
ollama.generate("./llama-7b-ggml.bin", "hi")

ollama.generate(model, message)

Generate a completion

ollama.generate("./llama-7b-ggml.bin", "hi")

ollama.load(model)

Load a model for generation

ollama.load("model")

ollama.models()

List available local models

models = ollama.models()

ollama.serve()

Serve the ollama http server

Cooming Soon

ollama.pull(model)

Download a model

ollama.pull("huggingface.co/thebloke/llama-7b-ggml")

ollama.import(filename)

Import a model from a file

ollama.import("./path/to/model")

ollama.search("query")

Search for compatible models that Ollama can run

ollama.search("llama-7b")

Future CLI

In the future, there will be an ollama CLI for running models on servers, in containers or for local development environments.

ollama generate huggingface.co/thebloke/llama-7b-ggml "hi"
> Downloading [================>          ] 66.67% (2/3) 30.2MB/s

Documentation

• Development