Added evaluation metrics
Browse files
README.md
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
pipeline_tag: text-generation
|
| 6 |
+
tags:
|
| 7 |
+
- nlp
|
| 8 |
+
- code
|
| 9 |
+
- microsoft
|
| 10 |
+
---
|
| 11 |
+
Official [AQLM](https://arxiv.org/abs/2401.06118) quantization of [microsoft/Phi-3-mini-128k-instruct
|
| 12 |
+
](https://huggingface.co/microsoft/Phi-3-mini-128k-instruct).
|
| 13 |
+
|
| 14 |
+
For this quantization, we used 1 codebook of 16 bits.
|
| 15 |
+
|
| 16 |
+
Results:
|
| 17 |
+
| Model | Quantization | MMLU (5-shot) | ArcC| ArcE| Hellaswag | Winogrande | PiQA | Model size, Gb |
|
| 18 |
+
|------|------|-------|------|------|------|------|------|------|
|
| 19 |
+
| microsoft/Phi-3-mini-128k-instruct| None | 0.6881 | 0.5418 | 0.8127 | 0.5980 | 0.7873 | 0.7340 | 7.6 |
|
| 20 |
+
| | 1x16 | 0.5815 | 0.4599 | 0.7845 | 0.5235 | 0.7666 | 0.6930 | 1.4 |
|