Commit
•
eb89aba
1
Parent(s):
7fe0ab9
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
tags:
|
| 4 |
+
- llama
|
| 5 |
+
- facebook
|
| 6 |
+
- meta
|
| 7 |
+
- llama-3
|
| 8 |
+
- conversational
|
| 9 |
+
- text-generation-inference
|
| 10 |
+
---
|
| 11 |
+
Official [AQLM](https://arxiv.org/abs/2401.06118) quantization of [meta-llama/Meta-Llama-3.1-8B
|
| 12 |
+
](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B) finetuned with [PV-Tuning](https://arxiv.org/abs/2405.14852).
|
| 13 |
+
|
| 14 |
+
For this quantization, we used 1 codebook of 16 bits and groupsize of 8.
|
| 15 |
+
|
| 16 |
+
Results:
|
| 17 |
+
| Model | Quantization | MMLU (5-shot) | ArcC| ArcE| Hellaswag | PiQA | Winogrande | Model size, Gb |
|
| 18 |
+
|------|------|-------|------|------|------|------|------|------|
|
| 19 |
+
|meta-llama/Meta-Llama-3.1-8B| None |0.6521 | 0.5145 | 0.8144 | 0.5998 | 0.8014 | 0.7356 | 16.1 |
|
| 20 |
+
| | 1x16g8 | 0.5524 | 0.4437 | 0.7765 | 0.5603 | 0.7797 | 0.6851 | 4.1 |
|
| 21 |
+
|
| 22 |
+
**Note**
|
| 23 |
+
|
| 24 |
+
We used `lm-eval=0.4.0` for evaluation.
|