Awesome

Python Bindings for llama.cpp

Simple Python bindings for @ggerganov's llama.cpp library. This package provides:

• Low-level access to C API via ctypes interface.
• High-level Python API for text completion
  • OpenAI-like API
  • LangChain compatibility
  • LlamaIndex compatibility
• OpenAI compatible web server
  • Local Copilot replacement
  • Function Calling support
  • Vision API support
  • Multiple Models

Documentation is available at https://llama-cpp-python.readthedocs.io/en/latest.

Installation

Requirements:

• Python 3.8+
• C compiler
  • Linux: gcc or clang
  • Windows: Visual Studio or MinGW
  • MacOS: Xcode

To install the package, run:

pip install llama-cpp-python

This will also build llama.cpp from source and install it alongside this python package.

If this fails, add --verbose to the pip install see the full cmake build log.

Pre-built Wheel (New)

It is also possible to install a pre-built wheel with basic CPU support.

pip install llama-cpp-python \
  --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu

Installation Configuration

llama.cpp supports a number of hardware acceleration backends to speed up inference as well as backend specific options. See the llama.cpp README for a full list.

All llama.cpp cmake build options can be set via the CMAKE_ARGS environment variable or via the --config-settings / -C cli flag during installation.

# Linux and Mac
CMAKE_ARGS="-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS" \
  pip install llama-cpp-python

# Windows
$env:CMAKE_ARGS = "-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS"
pip install llama-cpp-python

They can also be set via pip install -C / --config-settings command and saved to a requirements.txt file:

pip install --upgrade pip # ensure pip is up to date
pip install llama-cpp-python \
  -C cmake.args="-DGGML_BLAS=ON;-DGGML_BLAS_VENDOR=OpenBLAS"

# requirements.txt

llama-cpp-python -C cmake.args="-DGGML_BLAS=ON;-DGGML_BLAS_VENDOR=OpenBLAS"

Supported Backends

Below are some common backends, their build commands and any additional environment variables required.

To install with OpenBLAS, set the GGML_BLAS and GGML_BLAS_VENDOR environment variables before installing:

CMAKE_ARGS="-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS" pip install llama-cpp-python

To install with CUDA support, set the GGML_CUDA=on environment variable before installing:

CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python

Pre-built Wheel (New)

It is also possible to install a pre-built wheel with CUDA support. As long as your system meets some requirements:

• CUDA Version is 12.1, 12.2, 12.3, 12.4 or 12.5
• Python Version is 3.10, 3.11 or 3.12

pip install llama-cpp-python \
  --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/<cuda-version>

Where <cuda-version> is one of the following:

• cu121: CUDA 12.1
• cu122: CUDA 12.2
• cu123: CUDA 12.3
• cu124: CUDA 12.4
• cu125: CUDA 12.5

For example, to install the CUDA 12.1 wheel:

pip install llama-cpp-python \
  --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu121

To install with Metal (MPS), set the GGML_METAL=on environment variable before installing:

CMAKE_ARGS="-DGGML_METAL=on" pip install llama-cpp-python

Pre-built Wheel (New)

It is also possible to install a pre-built wheel with Metal support. As long as your system meets some requirements:

• MacOS Version is 11.0 or later
• Python Version is 3.10, 3.11 or 3.12

pip install llama-cpp-python \
  --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/metal

To install with hipBLAS / ROCm support for AMD cards, set the GGML_HIPBLAS=on environment variable before installing:

CMAKE_ARGS="-DGGML_HIPBLAS=on" pip install llama-cpp-python

To install with Vulkan support, set the GGML_VULKAN=on environment variable before installing:

CMAKE_ARGS="-DGGML_VULKAN=on" pip install llama-cpp-python

To install with SYCL support, set the GGML_SYCL=on environment variable before installing:

source /opt/intel/oneapi/setvars.sh   
CMAKE_ARGS="-DGGML_SYCL=on -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx" pip install llama-cpp-python

To install with RPC support, set the GGML_RPC=on environment variable before installing:

source /opt/intel/oneapi/setvars.sh   
CMAKE_ARGS="-DGGML_RPC=on" pip install llama-cpp-python

Windows Notes

If you run into issues where it complains it can't find 'nmake' '?' or CMAKE_C_COMPILER, you can extract w64devkit as mentioned in llama.cpp repo and add those manually to CMAKE_ARGS before running pip install:

$env:CMAKE_GENERATOR = "MinGW Makefiles"
$env:CMAKE_ARGS = "-DGGML_OPENBLAS=on -DCMAKE_C_COMPILER=C:/w64devkit/bin/gcc.exe -DCMAKE_CXX_COMPILER=C:/w64devkit/bin/g++.exe"

See the above instructions and set CMAKE_ARGS to the BLAS backend you want to use.

MacOS Notes

Detailed MacOS Metal GPU install documentation is available at docs/install/macos.md

Note: If you are using Apple Silicon (M1) Mac, make sure you have installed a version of Python that supports arm64 architecture. For example:

wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-MacOSX-arm64.sh
bash Miniforge3-MacOSX-arm64.sh

Otherwise, while installing it will build the llama.cpp x86 version which will be 10x slower on Apple Silicon (M1) Mac.

Try installing with

CMAKE_ARGS="-DCMAKE_OSX_ARCHITECTURES=arm64 -DCMAKE_APPLE_SILICON_PROCESSOR=arm64 -DGGML_METAL=on" pip install --upgrade --verbose --force-reinstall --no-cache-dir llama-cpp-python

Upgrading and Reinstalling

To upgrade and rebuild llama-cpp-python add --upgrade --force-reinstall --no-cache-dir flags to the pip install command to ensure the package is rebuilt from source.

High-level API

API Reference

The high-level API provides a simple managed interface through the Llama class.

Below is a short example demonstrating how to use the high-level API to for basic text completion:

from llama_cpp import Llama

llm = Llama(
      model_path="./models/7B/llama-model.gguf",
      # n_gpu_layers=-1, # Uncomment to use GPU acceleration
      # seed=1337, # Uncomment to set a specific seed
      # n_ctx=2048, # Uncomment to increase the context window
)
output = llm(
      "Q: Name the planets in the solar system? A: ", # Prompt
      max_tokens=32, # Generate up to 32 tokens, set to None to generate up to the end of the context window
      stop=["Q:", "\n"], # Stop generating just before the model would generate a new question
      echo=True # Echo the prompt back in the output
) # Generate a completion, can also call create_completion
print(output)

By default llama-cpp-python generates completions in an OpenAI compatible format:

{
  "id": "cmpl-xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "object": "text_completion",
  "created": 1679561337,
  "model": "./models/7B/llama-model.gguf",
  "choices": [
    {
      "text": "Q: Name the planets in the solar system? A: Mercury, Venus, Earth, Mars, Jupiter, Saturn, Uranus, Neptune and Pluto.",
      "index": 0,
      "logprobs": None,
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 14,
    "completion_tokens": 28,
    "total_tokens": 42
  }
}

Text completion is available through the __call__ and create_completion methods of the Llama class.

Pulling models from Hugging Face Hub

You can download Llama models in gguf format directly from Hugging Face using the from_pretrained method. You'll need to install the huggingface-hub package to use this feature (pip install huggingface-hub).

llm = Llama.from_pretrained(
    repo_id="Qwen/Qwen2-0.5B-Instruct-GGUF",
    filename="*q8_0.gguf",
    verbose=False
)

By default from_pretrained will download the model to the huggingface cache directory, you can then manage installed model files with the huggingface-cli tool.

Chat Completion

The high-level API also provides a simple interface for chat completion.

Chat completion requires that the model knows how to format the messages into a single prompt. The Llama class does this using pre-registered chat formats (ie. chatml, llama-2, gemma, etc) or by providing a custom chat handler object.

The model will will format the messages into a single prompt using the following order of precedence:

• Use the chat_handler if provided
• Use the chat_format if provided
• Use the tokenizer.chat_template from the gguf model's metadata (should work for most new models, older models may not have this)
• else, fallback to the llama-2 chat format

Set verbose=True to see the selected chat format.

from llama_cpp import Llama
llm = Llama(
      model_path="path/to/llama-2/llama-model.gguf",
      chat_format="llama-2"
)
llm.create_chat_completion(
      messages = [
          {"role": "system", "content": "You are an assistant who perfectly describes images."},
          {
              "role": "user",
              "content": "Describe this image in detail please."
          }
      ]
)

Chat completion is available through the create_chat_completion method of the Llama class.

For OpenAI API v1 compatibility, you use the create_chat_completion_openai_v1 method which will return pydantic models instead of dicts.

JSON and JSON Schema Mode

To constrain chat responses to only valid JSON or a specific JSON Schema use the response_format argument in create_chat_completion.

JSON Mode

The following example will constrain the response to valid JSON strings only.

from llama_cpp import Llama
llm = Llama(model_path="path/to/model.gguf", chat_format="chatml")
llm.create_chat_completion(
    messages=[
        {
            "role": "system",
            "content": "You are a helpful assistant that outputs in JSON.",
        },
        {"role": "user", "content": "Who won the world series in 2020"},
    ],
    response_format={
        "type": "json_object",
    },
    temperature=0.7,
)

JSON Schema Mode

To constrain the response further to a specific JSON Schema add the schema to the schema property of the response_format argument.

from llama_cpp import Llama
llm = Llama(model_path="path/to/model.gguf", chat_format="chatml")
llm.create_chat_completion(
    messages=[
        {
            "role": "system",
            "content": "You are a helpful assistant that outputs in JSON.",
        },
        {"role": "user", "content": "Who won the world series in 2020"},
    ],
    response_format={
        "type": "json_object",
        "schema": {
            "type": "object",
            "properties": {"team_name": {"type": "string"}},
            "required": ["team_name"],
        },
    },
    temperature=0.7,
)

Function Calling

The high-level API supports OpenAI compatible function and tool calling. This is possible through the functionary pre-trained models chat format or through the generic chatml-function-calling chat format.

from llama_cpp import Llama
llm = Llama(model_path="path/to/chatml/llama-model.gguf", chat_format="chatml-function-calling")
llm.create_chat_completion(
      messages = [
        {
          "role": "system",
          "content": "A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. The assistant calls functions with appropriate input when necessary"

        },
        {
          "role": "user",
          "content": "Extract Jason is 25 years old"
        }
      ],
      tools=[{
        "type": "function",
        "function": {
          "name": "UserDetail",
          "parameters": {
            "type": "object",
            "title": "UserDetail",
            "properties": {
              "name": {
                "title": "Name",
                "type": "string"
              },
              "age": {
                "title": "Age",
                "type": "integer"
              }
            },
            "required": [ "name", "age" ]
          }
        }
      }],
      tool_choice={
        "type": "function",
        "function": {
          "name": "UserDetail"
        }
      }
)

The various gguf-converted files for this set of models can be found here. Functionary is able to intelligently call functions and also analyze any provided function outputs to generate coherent responses. All v2 models of functionary supports parallel function calling. You can provide either functionary-v1 or functionary-v2 for the chat_format when initializing the Llama class.

Due to discrepancies between llama.cpp and HuggingFace's tokenizers, it is required to provide HF Tokenizer for functionary. The LlamaHFTokenizer class can be initialized and passed into the Llama class. This will override the default llama.cpp tokenizer used in Llama class. The tokenizer files are already included in the respective HF repositories hosting the gguf files.

from llama_cpp import Llama
from llama_cpp.llama_tokenizer import LlamaHFTokenizer
llm = Llama.from_pretrained(
  repo_id="meetkai/functionary-small-v2.2-GGUF",
  filename="functionary-small-v2.2.q4_0.gguf",
  chat_format="functionary-v2",
  tokenizer=LlamaHFTokenizer.from_pretrained("meetkai/functionary-small-v2.2-GGUF")
)

NOTE: There is no need to provide the default system messages used in Functionary as they are added automatically in the Functionary chat handler. Thus, the messages should contain just the chat messages and/or system messages that provide additional context for the model (e.g.: datetime, etc.).

Multi-modal Models

llama-cpp-python supports such as llava1.5 which allow the language model to read information from both text and images.

Below are the supported multi-modal models and their respective chat handlers (Python API) and chat formats (Server API).

Then you'll need to use a custom chat handler to load the clip model and process the chat messages and images.

from llama_cpp import Llama
from llama_cpp.llama_chat_format import Llava15ChatHandler
chat_handler = Llava15ChatHandler(clip_model_path="path/to/llava/mmproj.bin")
llm = Llama(
  model_path="./path/to/llava/llama-model.gguf",
  chat_handler=chat_handler,
  n_ctx=2048, # n_ctx should be increased to accommodate the image embedding
)
llm.create_chat_completion(
    messages = [
        {"role": "system", "content": "You are an assistant who perfectly describes images."},
        {
            "role": "user",
            "content": [
                {"type" : "text", "text": "What's in this image?"},
                {"type": "image_url", "image_url": {"url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg" } }
            ]
        }
    ]
)

You can also pull the model from the Hugging Face Hub using the from_pretrained method.

from llama_cpp import Llama
from llama_cpp.llama_chat_format import MoondreamChatHandler

chat_handler = MoondreamChatHandler.from_pretrained(
  repo_id="vikhyatk/moondream2",
  filename="*mmproj*",
)

llm = Llama.from_pretrained(
  repo_id="vikhyatk/moondream2",
  filename="*text-model*",
  chat_handler=chat_handler,
  n_ctx=2048, # n_ctx should be increased to accommodate the image embedding
)

response = llm.create_chat_completion(
    messages = [
        {
            "role": "user",
            "content": [
                {"type" : "text", "text": "What's in this image?"},
                {"type": "image_url", "image_url": {"url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg" } }

            ]
        }
    ]
)
print(response["choices"][0]["text"])

Note: Multi-modal models also support tool calling and JSON mode.

Images can be passed as base64 encoded data URIs. The following example demonstrates how to do this.

import base64

def image_to_base64_data_uri(file_path):
    with open(file_path, "rb") as img_file:
        base64_data = base64.b64encode(img_file.read()).decode('utf-8')
        return f"data:image/png;base64,{base64_data}"

# Replace 'file_path.png' with the actual path to your PNG file
file_path = 'file_path.png'
data_uri = image_to_base64_data_uri(file_path)

messages = [
    {"role": "system", "content": "You are an assistant who perfectly describes images."},
    {
        "role": "user",
        "content": [
            {"type": "image_url", "image_url": {"url": data_uri }},
            {"type" : "text", "text": "Describe this image in detail please."}
        ]
    }
]

Speculative Decoding

llama-cpp-python supports speculative decoding which allows the model to generate completions based on a draft model.

The fastest way to use speculative decoding is through the LlamaPromptLookupDecoding class.

Just pass this as a draft model to the Llama class during initialization.

from llama_cpp import Llama
from llama_cpp.llama_speculative import LlamaPromptLookupDecoding

llama = Llama(
    model_path="path/to/model.gguf",
    draft_model=LlamaPromptLookupDecoding(num_pred_tokens=10) # num_pred_tokens is the number of tokens to predict 10 is the default and generally good for gpu, 2 performs better for cpu-only machines.
)

Embeddings

To generate text embeddings use create_embedding or embed. Note that you must pass embedding=True to the constructor upon model creation for these to work properly.

import llama_cpp

llm = llama_cpp.Llama(model_path="path/to/model.gguf", embedding=True)

embeddings = llm.create_embedding("Hello, world!")

# or create multiple embeddings at once

embeddings = llm.create_embedding(["Hello, world!", "Goodbye, world!"])

There are two primary notions of embeddings in a Transformer-style model: token level and sequence level. Sequence level embeddings are produced by "pooling" token level embeddings together, usually by averaging them or using the first token.

Models that are explicitly geared towards embeddings will usually return sequence level embeddings by default, one for each input string. Non-embedding models such as those designed for text generation will typically return only token level embeddings, one for each token in each sequence. Thus the dimensionality of the return type will be one higher for token level embeddings.

It is possible to control pooling behavior in some cases using the pooling_type flag on model creation. You can ensure token level embeddings from any model using LLAMA_POOLING_TYPE_NONE. The reverse, getting a generation oriented model to yield sequence level embeddings is currently not possible, but you can always do the pooling manually.

Adjusting the Context Window

The context window of the Llama models determines the maximum number of tokens that can be processed at once. By default, this is set to 512 tokens, but can be adjusted based on your requirements.

For instance, if you want to work with larger contexts, you can expand the context window by setting the n_ctx parameter when initializing the Llama object:

llm = Llama(model_path="./models/7B/llama-model.gguf", n_ctx=2048)

OpenAI Compatible Web Server

llama-cpp-python offers a web server which aims to act as a drop-in replacement for the OpenAI API. This allows you to use llama.cpp compatible models with any OpenAI compatible client (language libraries, services, etc).

To install the server package and get started:

pip install 'llama-cpp-python[server]'
python3 -m llama_cpp.server --model models/7B/llama-model.gguf

Similar to Hardware Acceleration section above, you can also install with GPU (cuBLAS) support like this:

CMAKE_ARGS="-DGGML_CUDA=on" FORCE_CMAKE=1 pip install 'llama-cpp-python[server]'
python3 -m llama_cpp.server --model models/7B/llama-model.gguf --n_gpu_layers 35

Navigate to http://localhost:8000/docs to see the OpenAPI documentation.

To bind to 0.0.0.0 to enable remote connections, use python3 -m llama_cpp.server --host 0.0.0.0. Similarly, to change the port (default is 8000), use --port.

You probably also want to set the prompt format. For chatml, use

python3 -m llama_cpp.server --model models/7B/llama-model.gguf --chat_format chatml

That will format the prompt according to how model expects it. You can find the prompt format in the model card. For possible options, see llama_cpp/llama_chat_format.py and look for lines starting with "@register_chat_format".

If you have huggingface-hub installed, you can also use the --hf_model_repo_id flag to load a model from the Hugging Face Hub.

python3 -m llama_cpp.server --hf_model_repo_id Qwen/Qwen2-0.5B-Instruct-GGUF --model '*q8_0.gguf'

Web Server Features

• Local Copilot replacement
• Function Calling support
• Vision API support
• Multiple Models

Docker image

A Docker image is available on GHCR. To run the server:

docker run --rm -it -p 8000:8000 -v /path/to/models:/models -e MODEL=/models/llama-model.gguf ghcr.io/abetlen/llama-cpp-python:latest

Docker on termux (requires root) is currently the only known way to run this on phones, see termux support issue

Low-level API

API Reference

The low-level API is a direct ctypes binding to the C API provided by llama.cpp. The entire low-level API can be found in llama_cpp/llama_cpp.py and directly mirrors the C API in llama.h.

Below is a short example demonstrating how to use the low-level API to tokenize a prompt:

import llama_cpp
import ctypes
llama_cpp.llama_backend_init(False) # Must be called once at the start of each program
params = llama_cpp.llama_context_default_params()
# use bytes for char * params
model = llama_cpp.llama_load_model_from_file(b"./models/7b/llama-model.gguf", params)
ctx = llama_cpp.llama_new_context_with_model(model, params)
max_tokens = params.n_ctx
# use ctypes arrays for array params
tokens = (llama_cpp.llama_token * int(max_tokens))()
n_tokens = llama_cpp.llama_tokenize(ctx, b"Q: Name the planets in the solar system? A: ", tokens, max_tokens, llama_cpp.c_bool(True))
llama_cpp.llama_free(ctx)

Check out the examples folder for more examples of using the low-level API.

Documentation

Documentation is available via https://llama-cpp-python.readthedocs.io/. If you find any issues with the documentation, please open an issue or submit a PR.

Development

This package is under active development and I welcome any contributions.

To get started, clone the repository and install the package in editable / development mode:

git clone --recurse-submodules https://github.com/abetlen/llama-cpp-python.git
cd llama-cpp-python

# Upgrade pip (required for editable mode)
pip install --upgrade pip

# Install with pip
pip install -e .

# if you want to use the fastapi / openapi server
pip install -e '.[server]'

# to install all optional dependencies
pip install -e '.[all]'

# to clear the local build cache
make clean

Now try running the tests

pytest

There's a Makefile available with useful targets. A typical workflow would look like this:

make build
make test

You can also test out specific commits of llama.cpp by checking out the desired commit in the vendor/llama.cpp submodule and then running make clean and pip install -e . again. Any changes in the llama.h API will require changes to the llama_cpp/llama_cpp.py file to match the new API (additional changes may be required elsewhere).

FAQ

Are there pre-built binaries / binary wheels available?

The recommended installation method is to install from source as described above. The reason for this is that llama.cpp is built with compiler optimizations that are specific to your system. Using pre-built binaries would require disabling these optimizations or supporting a large number of pre-built binaries for each platform.

That being said there are some pre-built binaries available through the Releases as well as some community provided wheels.

In the future, I would like to provide pre-built binaries and wheels for common platforms and I'm happy to accept any useful contributions in this area. This is currently being tracked in #741

How does this compare to other Python bindings of llama.cpp?

I originally wrote this package for my own use with two goals in mind:

• Provide a simple process to install llama.cpp and access the full C API in llama.h from Python
• Provide a high-level Python API that can be used as a drop-in replacement for the OpenAI API so existing apps can be easily ported to use llama.cpp

Any contributions and changes to this package will be made with these goals in mind.

License

This project is licensed under the terms of the MIT license.