Commit
•
51ccd37
1
Parent(s):
7fe8e72
Added description and evaluation metrics
Browse files
README.md
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
- fr
|
| 5 |
+
- de
|
| 6 |
+
- es
|
| 7 |
+
- it
|
| 8 |
+
- pt
|
| 9 |
+
- ru
|
| 10 |
+
- zh
|
| 11 |
+
- ja
|
| 12 |
+
license: apache-2.0
|
| 13 |
+
---
|
| 14 |
+
Official [AQLM](https://arxiv.org/abs/2401.06118) quantization of [mistralai/Mistral-Nemo-Instruct-2407
|
| 15 |
+
](https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407) finetuned with [PV-Tuning](https://arxiv.org/abs/2405.14852).
|
| 16 |
+
|
| 17 |
+
For this quantization, we used 1 codebook of 16 bits and groupsize of 8.
|
| 18 |
+
|
| 19 |
+
Results:
|
| 20 |
+
| Model | Quantization | MMLU (5-shot) | ArcC| ArcE| Hellaswag | PiQA | Winogrande | Model size, Gb |
|
| 21 |
+
|------|------|-------|------|------|------|------|------|------|
|
| 22 |
+
|mistralai/Mistral-Nemo-Instruct-2407| None | 0.6819 | 0.5606 | 0.8241 | 0.6332 | 0.8090 | 0.7498 | 24.5 |
|
| 23 |
+
| | 1x16g8 | 0.6071 | 0.5017 | 0.7942 | 0.5930 | 0.7987 | 0.7356 | 5.4 |
|
| 24 |
+
|
| 25 |
+
**Note**
|
| 26 |
+
|
| 27 |
+
We used `lm-eval=0.4.0` for evaluation.
|