Skip to main content

mixtral.py

A Python module for running the Mixtral-8x7B language model with customisable precision and attention mechanisms. Mixtral-8x7B, being a pretrained base model, lacks moderation mechanisms. Read more here.

Installation

pip/pip3 install mixtral.py

Usage

Run the model

import mixtral
mixtral = Mixtral()
print("Running the model:")
print(mixtral.generate_text("..."))

In half-precision

import mixtral
print("Running the model in half precision:")
mixtral_half_precision = Mixtral(use_half_precision=True)
print(mixtral_half_precision.generate_text("..."))

Lower precision using (8-bit & 4-bit) using bitsandbytes

import mixtral
print("Running the model in lower precision using (8-bit & 4-bit) using bitsandbytes:")
mixtral_4bit = Mixtral(load_in_4bit=True)
print(mixtral_4bit.generate_text("..."))

Load the model with Flash Attention 2

import mixtral
print("Load the model with Flash Attention 2:")
mixtral_flash_attention_2 = Mixtral(use_flash_attention_2=True)
print(mixtral_flash_attention_2.generate_text("..."))

Hardware Requirements

  • minimum 100GB GPU RAM (Mistral AI)
  • Hardware requirements after fine-tuning can be found in the discussion here

Licenses

Contributing

I welcome contributions from the community and appreciate the time and effort put into making mixtral.py better. To contribute, please follow the guidelines and steps outlined below:

Note: Your pull request will be closed if you do not specify the changes you've made.

Fork the Repository

Start by forking this repository. You can do this by clicking on the "Fork" button located at the top right corner of the GitHub page. This will create a personal copy of the repository under your own GitHub account.

Clone the Repository

Next, clone the forked repository to your local machine using the following command:

$ git clone https://github.com/yourusername/mixtral.py.git

Navigate to the cloned directory:

$ cd mixtral.py

Create a New Branch

Before making any changes, it's recommended to create a new branch. This ensures that your changes won't interfere with other contributions and keeps the main branch clean. Use the following command to create and switch to a new branch:

$ git checkout -b branch-name

Make the Desired Changes

Now, you can proceed to make your desired changes to the project. Whether it's fixing bugs, adding new features, improving documentation, or optimizing code, your efforts will be instrumental in enhancing the project.

Commit and Push Changes

Once you have made the necessary changes, commit your work using the following commands:

$ git add .
$ git commit -m "Your commit message"

Push the changes to your forked repository:

$ git push origin branch-name

Submit a Pull Request

Head over to the original repository on GitHub and go to the "Pull requests" tab.

  1. Click on the "New pull request" button.
  2. Select your forked repository and the branch containing your changes.
  3. Provide a clear and informative title for your pull request, and use the description box to explain the modifications you have made. Your pull request will be closed if you do not specify the changes you've made.
  4. Finally, click on the "Create pull request" button to submit your changes.

Verifying Signatures

Import PGP Key into Keyring

gpg --keyserver 185.125.188.27 --recv-keys 20247EC023F2769E66181C0F581B4A2A862BBADE

or

wget https://github.com/ibnaleem/ibnaleem/blob/main/public_key.asc

Download Signature

The signatures (.asc and .sig) can be found in the /sig directory. Download either of them. Open an issue with the title "invalid signature/wrong signature/corrupt signature" for any issues regarding my signatures.

Sign My Key

gpg --sign-key 20247EC023F2769E66181C0F581B4A2A862BBADE
gpg --send-keys 20247EC023F2769E66181C0F581B4A2A862BBADE

If you cannot upload your signature, see "gpg: keyserver receive failed: No route to host"

Verify

gpg --verify mixtral.py.asc mixtral.py

Desired output:

gpg: Signature made Tue  6 Feb 10:27:34 2024 GMT
gpg:                using RSA key 20247EC023F2769E66181C0F581B4A2A862BBADE
gpg: Good signature from "Ibn Aleem <ibnaleem@outlook.com>" [ultimate]

Acknowledgements

Release files for mixtral.py 0.0.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for mixtral.py 0.0.1
File Size Uploaded
mixtral.py-0.0.1.tar.gz 15.6 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for mixtral.py 0.0.1
File Interpreter ABI Platform
mixtral.py-0.0.1-py3-none-any.whl Python 3 none any Details

Total release size: 31.2 kB

Release files / mixtral.py-0.0.1.tar.gz

Download URL mixtral.py-0.0.1.tar.gz
Size 15.6 kB
Tags Source
SHA-256 checksum
How to use checksums
39341daaece0a0fe11e9fcb5911c6ce41030c4d2213be7b7889888f9dc4c782a
BLAKE2b-256 checksum
How to use checksums
55b82005a6c4f3f9aa86771bdef8eb5fee77cb1b027a2ccaf0a5e6e5e824b631
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/4.0.2 CPython/3.11.5

Release files / mixtral.py-0.0.1-py3-none-any.whl

Download URL mixtral.py-0.0.1-py3-none-any.whl
Size 15.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
2a39e76a271a43da43d5b50402a4fe2a9b0f8038fd28b07b4b76531135292cb5
BLAKE2b-256 checksum
How to use checksums
ca1c2493ec677dc53e7c008ace693f3903b8c495ecbac0257baf8abedee2d05a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/4.0.2 CPython/3.11.5

Release history Release notifications | RSS feed

This release

0.0.1 This release

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page