cwm-q4-mlx / README.md
bradeenl's picture
Upload folder using huggingface_hub
f502aeb verified
metadata
language:
  - en
tags:
  - meta-ai
  - meta-pytorch
license: fair-noncommercial-research-license
license-link: https://huggingface.co/facebook/fair-noncommercial-research-license

Code World Model (CWM)

This repository contains the MLX conversion of CWM with a 4bit quantization.

mlx-lm mlx_lm.generate --model . --prompt "hello"                                 
==========
Hello! It's nice to meet you. Is there something I can help you with or would you like to chat?
</think>
==========
Prompt: 14 tokens, 20.107 tokens-per-sec
Generation: 28 tokens, 18.347 tokens-per-sec
Peak memory: 18.408 GB