|
--- |
|
library_name: transformers |
|
tags: |
|
- mistral |
|
- finetuned |
|
- conversational |
|
- text-generation-inference |
|
--- |
|
Official [AQLM](https://arxiv.org/abs/2401.06118) quantization of [mistralai/Mistral-7B-Instruct-v0.2 |
|
](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2). |
|
|
|
For this quantization, we used 2 codebooks of 8 bits. |
|
|
|
Results: |
|
| Model | Quantization | MMLU (5-shot) | Model size, Gb | |
|
|------|------|------|------| |
|
|mistralai/Mistral-7B-Instruct-v0.2 | None | 0.5912 | 14.5 | |
|
| | 2x8 | 0.4384 | 2.3 | |
