|
|
--- |
|
|
license: apache-2.0 |
|
|
pipeline_tag: text-generation |
|
|
library_name: transformers |
|
|
tags: |
|
|
- vllm |
|
|
- mlx |
|
|
- quantized |
|
|
base_model: openai/gpt-oss-120b |
|
|
--- |
|
|
|
|
|
# gpt-oss-120B (6‑bit quantized via MLX‑LM) |
|
|
|
|
|
**A 6‑bit quantized version of `openai/gpt-oss-120b` created using MLX‑LM.** |
|
|
This version significantly reduces inference memory requirements (~90 GB RAM), while retaining most of the model’s original capabilities. |
|
|
|
|
|
⸻ |
|
|
|
|
|
🛠️ Quantization Process |
|
|
|
|
|
The model was created using the following steps: |
|
|
|
|
|
* pip uninstall mlx-lm |
|
|
* pip install git+https://github.com/ml-explore/mlx-lm.git@main |
|
|
* mlx_lm.convert \ |
|
|
--hf-path openai/gpt-oss-120b \ |
|
|
--quantize \ |
|
|
--q-bits 6 \ |
|
|
--output-dir gpt-oss-120b-MLX-6bit |
|
|
|
|
|
These commands use the latest MLX‑LM converter to apply a consistent 6‑bit quantization across model weights. |
