Commit
•
2bcaf27
1
Parent(s):
cd9cabb
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
tags:
|
| 4 |
+
- phi-3
|
| 5 |
+
- phi-3-mini
|
| 6 |
+
- phi-3-mini-4k-instruct
|
| 7 |
+
- conversational
|
| 8 |
+
- text-generation-inference
|
| 9 |
+
pipeline_tag: text-generation
|
| 10 |
+
language:
|
| 11 |
+
- en
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
Official quantization of [Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) using [PV-Tuning](https://arxiv.org/abs/2405.14852) on top of [AQLM](https://arxiv.org/abs/2401.06118).
|
| 15 |
+
|
| 16 |
+
For this quantization, we used 1 codebook of 16 bits for groups of 8 weights.
|
| 17 |
+
|
| 18 |
+
Results (0-shot `acc`):
|
| 19 |
+
|
| 20 |
+
Results:
|
| 21 |
+
| Model | Quantization | ArcC| ArcE| Hellaswag | PiQA | Winogrande | Model size, Gb |
|
| 22 |
+
|------|------|-------|------|------|------|------|------|------|
|
| 23 |
+
| microsoft/Phi-3-mini-4k-instruct| None | 0.5529 | 0.8325 | 0.6055 | 0.8020 | 0.7364 | 7.6 |
|
| 24 |
+
| | 1x16 | 0.5051 | 0.7950 | 0.5532 | 0.7949 | 73.01 | 1.4 |
|
| 25 |
+
|
| 26 |
+
The 1x16g16 (1-bit) models are on the way, as soon as we update the inference lib with their respective kernels.
|
| 27 |
+
|
| 28 |
+
To learn more about the inference, as well as the information on how to quantize models yourself, please refer to the [official GitHub repo](https://github.com/Vahe1994/AQLM).
|
| 29 |
+
The original code for PV-Tuning can be found in the [AQLM@pv-tuning](https://github.com/Vahe1994/AQLM/tree/pv-tuning) branch.
|