Hugging Face's logo

Skywork
/
Skywork-R1V-38B-AWQ

Image-Text-to-Text
Transformers
PyTorch
Model card Files
xet
 Community
3
OrlandoHugBot commited on
Commit
0fd378a
·
verified ·
1 Parent(s): f966d33

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +73 -21
README.md CHANGED
@@ -13,21 +13,84 @@
13
  </div>
14
 
15
 
16
- ## Benchmark Results
17
 
18
- The AWQ quantized model maintains strong performance across key benchmarks:
19
-
20
- | Benchmark | Score |
21
- |-----------|-------|
22
- | MMMU | 0.6 |
23
- | MathV | 0.59 |
24
- | AIME_2024 | 0.6 |
25
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26
 
27
  ## Usage
28
-
29
  You can use the quantized model with different inference frameworks:
30
-
31
  ### Using VLLM
32
 
33
  #### Python API
@@ -111,19 +174,8 @@ If you use this model in your research, please cite:
111
 
112
  # Skywork-R1V-38B-AWQ (中文说明)
113
 
114
- ## 基准测试结果
115
-
116
- AWQ 量化模型在关键基准测试中保持了强劲的性能：
117
-
118
- | 基准测试 | 分数 |
119
- |-----------|-------|
120
- | MMMU | 0.6 |
121
- | MathV | 0.59 |
122
- | AIME_2024 | 0.6 |
123
-
124
 
125
  ## 使用方法
126
-
127
  您可以使用不同的推理框架来使用这个量化模型：
128
 
129
  ### 使用 VLLM
 
13
  </div>
14
 
15
 
16
+ ## Evaluation
17
 
18
+ <div align="center">
19
+ <b>Comparison with Larger-Scale Open-Source and Closed-Source Models</b>
20
+ </div>
 
 
 
 
21
 
22
+ <table align="center">
23
+ <thead>
24
+ <tr>
25
+ <th></th>
26
+ <th align="center"><strong>Benchmark</strong></th>
27
+ <th align="center"><strong>LLM</strong></th>
28
+ <th align="center" colspan="5"><strong>VLM</strong></th>
29
+ </tr>
30
+ <tr>
31
+ <th></th>
32
+ <th></th>
33
+ <th align="center"><strong>QwQ-32B-Preview</strong></th>
34
+ <th align="center"><strong>InternVL-2.5-38B</strong></th>
35
+ <th align="center"><strong>VILA 1.5-40B</strong></th>
36
+ <th align="center"><strong>InternVL2-40B</strong></th>
37
+ <th align="center"><strong>Skywork-R1V-38B</strong></th>
38
+ <th align="center"><strong>Skywork-R1V-AWQ</strong></th>
39
+ </tr>
40
+ </thead>
41
+ <tbody>
42
+ <tr>
43
+ <td rowspan="3">Reasoning</td>
44
+ <td>MATH-500</td>
45
+ <td align="center">90.6</td>
46
+ <td align="center">-</td>
47
+ <td align="center">-</td>
48
+ <td align="center">-</td>
49
+ <td align="center"><strong>94.0</strong></td>
50
+ <td align="center">86.0</td>
51
+ </tr>
52
+ <tr>
53
+ <td>AIME 2024</td>
54
+ <td align="center">50.0</td>
55
+ <td align="center">-</td>
56
+ <td align="center">-</td>
57
+ <td align="center">-</td>
58
+ <td align="center"><strong>72.0</strong></td>
59
+ <td align="center">61.0</td>
60
+ </tr>
61
+ <tr>
62
+ <td>GPQA</td>
63
+ <td align="center">54.5</td>
64
+ <td align="center">-</td>
65
+ <td align="center">-</td>
66
+ <td align="center">-</td>
67
+ <td align="center"><strong>61.6</strong></td>
68
+ <td align="center">56.5</td>
69
+ </tr>
70
+ <tr>
71
+ <td rowspan="2">Vision</td>
72
+ <td>MathVista(mini)</td>
73
+ <td align="center">-</td>
74
+ <td align="center">71.9</td>
75
+ <td align="center">49.5</td>
76
+ <td align="center">63.7</td>
77
+ <td align="center">67.5</td>
78
+ <td align="center">59.9</td>
79
+ </tr>
80
+ <tr>
81
+ <td>MMMU(Val)</td>
82
+ <td align="center">-</td>
83
+ <td align="center">63.9</td>
84
+ <td align="center">55.1</td>
85
+ <td align="center">55.2</td>
86
+ <td align="center"><strong>69.0</strong></td>
87
+ <td align="center">60.1</td>
88
+ </tr>
89
+ </tbody>
90
+ </table>
91
 
92
  ## Usage
 
93
  You can use the quantized model with different inference frameworks:
 
94
  ### Using VLLM
95
 
96
  #### Python API
 
174
 
175
  # Skywork-R1V-38B-AWQ (中文说明)
176
 
 
 
 
 
 
 
 
 
 
 
177
 
178
  ## 使用方法
 
179
  您可以使用不同的推理框架来使用这个量化模型：
180
 
181
  ### 使用 VLLM