File size: 914 Bytes
eb89aba
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bcb391a
eb89aba
 
 
bcb391a
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
---
library_name: transformers
tags:
- llama
- facebook
- meta
- llama-3
- conversational
- text-generation-inference
---
Official [AQLM](https://arxiv.org/abs/2401.06118) quantization of [meta-llama/Meta-Llama-3.1-8B
](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B) finetuned with [PV-Tuning](https://arxiv.org/abs/2405.14852).

For this quantization, we used 1 codebook of 16 bits and groupsize of 8.

Results: 
| Model      | Quantization | MMLU (5-shot) | ArcC| ArcE| Hellaswag | PiQA | Winogrande | Model size, Gb |
|------|------|-------|------|------|------|------|------|------|
|meta-llama/Meta-Llama-3.1-8B| None |0.6521	| 0.5145 | 0.8144 | 0.5998 | 0.8014	| 0.7356 | 16.1 |
|  | 1x16g8 | 0.5746 | 0.4454 | 0.7744	| 0.5629 | 0.7954 | 0.7024 | 4.1 |

**Note**

We used `lm-eval=0.4.0` for evaluation.

**UPD (09.08.2024)**

Uploaded new version finetuned on more data for longer with better quality.