README.md

Mistral NeMo

Language: Python

API: MAX Graph

This pipeline provides optimized support for the MistralForCausalLM family of large language models, as exemplified by the Mistral NeMo 12B text completion model. The model itself has been constructed in Python using the MAX Graph API.

The MAX Graph API provides an accessible interface to the construction of flexible accelerated compute graphs, which are then optimized by the MAX Engine's advanced graph compiler. This pipeline showcases how a large language model can be fully defined using Python and MAX Graphs and then compiled for optimal inference performance via the MAX Engine.

Model

Mistral NeMo is an open source large language model released by Mistral AI in partnership with NVIDIA. The weight checkpoints have been released under an Apache 2.0 license.

The default settings for this pipeline use the NeMo 12B set of pretrained weights in the bfloat16 encoding.

Usage

The easiest way to try out this pipeline is with our Magic command-line tool.

1. Install Magic on macOS and Ubuntu with this command:

   curl -ssL https://magic.modular.com | bash

   Then run the source command that's printed in your terminal.

   To see the available commands, you can run magic --help. Learn more about Magic here.

2. Clone the MAX examples repository:

   If you don't already have a local clone of this repository, create one via:

   git clone https://github.com/modularml/max.git

   The following instructions assume that you're present within this pipeline's directory, and you can change to it after cloning:

   cd max/pipelines/python/

3. Now run the Mistral text completion demo with the following command:

   magic run mistral --prompt "I believe the meaning of life is"

4. Host a chat completion endpoint via MAX Serve.

   MAX Serve provides functionality to host performant OpenAI compatible endpoints using the FastAPI framework.

   You can configure the pipeline to be hosted by using the --serve argument. For example:

   magic run mistral --serve

   A request can be submitted via a cURL command.

   curl -N http://localhost:8000/v1/chat/completions \
   -H "Content-Type: application/json" \
   -d '{
       "model": "mistralai/Mistral-Nemo-Instruct-2407",
       "stream": true,
       "messages": [
           {"role": "system", "content": "You are a helpful assistant."},
           {"role": "user", "content": "Who won the world series in 2020?"}
       ]
   }'

   Additionally, finetuned weights hosted on Hugging Face for any compatible MistralForCausalLM model can be used with this optimized architecture via the serve command:

   magic run serve --huggingface-repo-id=mistralai/Ministral-8B-Instruct-2410

Options

The following command-line options are available to customize operation of the pipeline:

• --max-length: Controls the maximum length of the text sequence (includes the input tokens). (Default value: 512)
• --max-new-tokens: The maximum number of new tokens to generate. If a -1 value is provided, the model will continue to generate tokens for the entire context length. (Default value: -1)
• --prompt: The text prompt to use for further generation.
• --quantization-encoding: The encoding to use for a datatype that can be quantized to a low bits per weight format. The options for quantized formats will download and cache default weights, but float32 requires the use of --weight-path to specify locally downloaded full-precision weights for use in the model. Valid values: bfloat16, float32. (Default value: float32).
• --save-to-serialized-model-path: If specified, writes the serialized model to this path.
• --serialized-model-path: If specified, tries to load a serialized model from this path.
• --top-k: Limits the sampling to the K most probable tokens. Default is 1.
• --weight-path: Overrides the default URL, and allows for an already-downloaded pretrained weight file to be used with the model.
• --max-cache-batch-size: Specifies the maximum batch size to be used. Default is 1.
• --use-gpu: Uses the GPU to execute the model.