Skip to content

Commit e80249b

Browse files
authored
update readme (#124)
1 parent 554cc80 commit e80249b

2 files changed

Lines changed: 32 additions & 6 deletions

File tree

README.md

Lines changed: 16 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -456,11 +456,24 @@ Benchmark results for DeepSeek-R1-0528 series models with `FP8-Block-Wise` and `
456456
457457
</details>
458458
459-
#### 1.4 Qwen-VL Series Models**
459+
#### 1.4 Qwen-VL Series Models
460460
461-
- **Qwen3-VL Benchmark**
461+
**Qwen3-VL Benchmark**
462462
463-
Coming soon.
463+
Benchmark results for Qwen3VL series models with `BF16`、`FP8-Static` and `FP8-Dynamic` quantization algorithms on datasets including `MMMU_VAL`、`DocVQA_VAL` and `ChartQA_TEST`:
464+
465+
<table>
466+
<thead>
467+
<tr><th>Model</th><th>Quantization</th><th>MMMU_VAL</th><th>DocVQA_VAL</th><th>ChartQA_TEST</th></tr>
468+
</thead>
469+
<tbody>
470+
<tr><td rowspan="3">Qwen3-VL-32B-Instruct</td><td>BF16</td><td>60.11</td><td>96.08</td><td>94.64</td></tr>
471+
<tr><td>FP8-Static</td><td>61.22</td><td>96.00</td><td>94.64</td></tr>
472+
<tr><td>FP8-Dynamic</td><td>60.78</td><td>96.19</td><td>94.72</td></tr>
473+
<tr><td rowspan="2">Qwen3-VL-30B-A3B-Instruct</td><td>BF16</td><td>50.44</td><td>95.28</td><td>95.36</td></tr>
474+
<tr><td>FP8-Dynamic</td><td>50.67</td><td>95.25</td><td>95.20</td></tr>
475+
</tbody>
476+
</table>
464477
465478
<details>
466479
<summary><strong>Qwen2.5VL Benchmark</strong></summary>

README_cn.md

Lines changed: 16 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -465,12 +465,25 @@ DeepSeek-R1-0528模型的`FP8-Block-Wise`、`W4A8-FP8`在`GPQA Diamond`、`AIME
465465
466466
#### 1.4 Qwen-VL 系列模型
467467
468-
- **Qwen3-VL Benchmark**
468+
**Qwen3-VL Benchmark**
469469
470-
待更新
470+
Qwen3VL系列模型的`BF16`、`FP8-Static`、`FP8-Dynamic`在`MMMU_VAL`、`DocVQA_VAL`、`ChartQA_TEST`上的评测结果如下:
471+
472+
<table>
473+
<thead>
474+
<tr><th>Model</th><th>Quantization</th><th>MMMU_VAL</th><th>DocVQA_VAL</th><th>ChartQA_TEST</th></tr>
475+
</thead>
476+
<tbody>
477+
<tr><td rowspan="3">Qwen3-VL-32B-Instruct</td><td>BF16</td><td>60.11</td><td>96.08</td><td>94.64</td></tr>
478+
<tr><td>FP8-Static</td><td>61.22</td><td>96.00</td><td>94.64</td></tr>
479+
<tr><td>FP8-Dynamic</td><td>60.78</td><td>96.19</td><td>94.72</td></tr>
480+
<tr><td rowspan="2">Qwen3-VL-30B-A3B-Instruct</td><td>BF16</td><td>50.44</td><td>95.28</td><td>95.36</td></tr>
481+
<tr><td>FP8-Dynamic</td><td>50.67</td><td>95.25</td><td>95.20</td></tr>
482+
</tbody>
483+
</table>
471484
472485
<details>
473-
<summary><li><strong>Qwen2.5VL Benchmark</strong></li></summary>
486+
<summary><strong>Qwen2.5VL Benchmark</strong></summary>
474487
475488
Qwen2.5VL系列模型的`BF16`、`FP8-Static`、`FP8-Dynamic`、`INT4-GPTQ`、`INT4-AWQ`在`MMMU_VAL`、`DocVQA_VAL`、`ChartQA_TEST`上的评测结果如下:
476489

0 commit comments

Comments
 (0)