Update model card
Browse files
README.md
CHANGED
|
@@ -6,4 +6,29 @@ base_model_relation: quantized
|
|
| 6 |
library_name: transformers
|
| 7 |
tags:
|
| 8 |
- qwq
|
| 9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
library_name: transformers
|
| 7 |
tags:
|
| 8 |
- qwq
|
| 9 |
+
- fp8
|
| 10 |
+
---
|
| 11 |
+
# Model Overview
|
| 12 |
+
|
| 13 |
+
## Description
|
| 14 |
+
|
| 15 |
+
FP8 Quantized QwQ-32B.
|
| 16 |
+
|
| 17 |
+
## Evaluation
|
| 18 |
+
|
| 19 |
+
The test results in the following table are based on the MMLU benchmark.
|
| 20 |
+
|
| 21 |
+
In order to speed up the test, we prevent the model from generating too long thought chains, so the score may be different from that with longer thought chain.
|
| 22 |
+
|
| 23 |
+
In our experiment, **the accuracy of the FP8 quantized version is almost the same as the BF16 version, and it can be used for faster inference.**
|
| 24 |
+
|
| 25 |
+
| Data Format | MMLU Score |
|
| 26 |
+
|:---|:---|
|
| 27 |
+
| BF16 Official | 61.2 |
|
| 28 |
+
| FP8 Quantized | 61.2 |
|
| 29 |
+
| Q8_0 (INT8) | 59.1 |
|
| 30 |
+
| AWQ (INT4) | 53.4 |
|
| 31 |
+
|
| 32 |
+
## Model Card Contact
|
| 33 |
+
|
| 34 |