Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
pipeline_tag: text-classification
|
3 |
+
---
|
4 |
+
|
5 |
+
## MiniCPM-V
|
6 |
+
**MiniCPM-V** is an efficient version with promising performance for deployment. The model is built based on MiniCPM-2.4B and SigLip-400M, connected by a perceiver resampler. Notable features of MiniCPM-V include:
|
7 |
+
|
8 |
+
- 🚀 **High Efficiency.**
|
9 |
+
|
10 |
+
MiniCPM-V can be **efficiently deployed on most GPU cards and personal computers**, and **even on edge devices such as mobile phones**. In terms of visual encoding, we compress the image representations into 64 tokens via a perceiver resampler, which is significantly fewer than other LMMs based on MLP architecture (typically > 512 tokens). This allows MiniCPM-V to operate with **much less memory cost and higher speed during inference**.
|
11 |
+
|
12 |
+
- 🔥 **Promising Performance.**
|
13 |
+
|
14 |
+
MiniCPM-V achieves **state-of-the-art performance** on multiple benchmarks (including MMMU, MME, and MMbech, etc) among models with comparable sizes, surpassing existing LMMs built on Phi-2. It even **achieves comparable or better performance than the 9.6B Qwen-VL-Chat**.
|
15 |
+
|
16 |
+
- 🙌 **Bilingual Support.**
|
17 |
+
|
18 |
+
MiniCPM-V is **the first edge-deployable LMM supporting bilingual multimodal interaction in English and Chinese**. This is achieved by generalizing multimodal capabilities across languages, a technique from our ICLR 2024 spotlight [paper](https://arxiv.org/abs/2308.12038).
|
19 |
+
|
20 |
+
<div align="center">
|
21 |
+
|
22 |
+
<table style="margin: 0px auto;">
|
23 |
+
<thead>
|
24 |
+
<tr>
|
25 |
+
<th align="left">Model</th>
|
26 |
+
<th>Size</th>
|
27 |
+
<th>MME</th>
|
28 |
+
<th nowrap="nowrap" >MMB dev (en)</th>
|
29 |
+
<th nowrap="nowrap" >MMB dev (zh)</th>
|
30 |
+
<th nowrap="nowrap" >MMMU val</th>
|
31 |
+
<th nowrap="nowrap" >CMMMU val</th>
|
32 |
+
</tr>
|
33 |
+
</thead>
|
34 |
+
<tbody align="center">
|
35 |
+
<tr>
|
36 |
+
<td align="left">LLaVA-Phi</td>
|
37 |
+
<td align="right">3.0B</td>
|
38 |
+
<td>1335</td>
|
39 |
+
<td>59.8</td>
|
40 |
+
<td>- </td>
|
41 |
+
<td>- </td>
|
42 |
+
<td>- </td>
|
43 |
+
</tr>
|
44 |
+
<tr>
|
45 |
+
<td nowrap="nowrap" align="left">MobileVLM</td>
|
46 |
+
<td align="right">3.0B</td>
|
47 |
+
<td>1289</td>
|
48 |
+
<td>59.6</td>
|
49 |
+
<td>- </td>
|
50 |
+
<td>- </td>
|
51 |
+
<td>- </td>
|
52 |
+
</tr>
|
53 |
+
<tr>
|
54 |
+
<td nowrap="nowrap" align="left" >Imp-v1</td>
|
55 |
+
<td align="right">3B</td>
|
56 |
+
<td>1434</td>
|
57 |
+
<td>66.5</td>
|
58 |
+
<td>- </td>
|
59 |
+
<td>- </td>
|
60 |
+
<td>- </td>
|
61 |
+
</tr>
|
62 |
+
<tr>
|
63 |
+
<td align="left" >Qwen-VL-Chat</td>
|
64 |
+
<td align="right" >9.6B</td>
|
65 |
+
<td>1487</td>
|
66 |
+
<td>60.6 </td>
|
67 |
+
<td>56.7 </td>
|
68 |
+
<td>35.9 </td>
|
69 |
+
<td>30.7 </td>
|
70 |
+
</tr>
|
71 |
+
<tr>
|
72 |
+
<td nowrap="nowrap" align="left" ><b>MiniCPM-V</b></td>
|
73 |
+
<td align="right">3B </td>
|
74 |
+
<td>1452 </td>
|
75 |
+
<td>67.3 </td>
|
76 |
+
<td>61.9 </td>
|
77 |
+
<td>34.7 </td>
|
78 |
+
<td>32.1 </td>
|
79 |
+
</tr>
|
80 |
+
</tbody>
|
81 |
+
</table>
|
82 |
+
|
83 |
+
</div>
|
84 |
+
|
85 |
+
<table>
|
86 |
+
<tr>
|
87 |
+
<td>
|
88 |
+
<p>
|
89 |
+
<img src="data/Mushroom_en.gif" width="400"/>
|
90 |
+
</p>
|
91 |
+
</td>
|
92 |
+
<td>
|
93 |
+
<p>
|
94 |
+
<img src="data/Snake_en.gif" width="400"/>
|
95 |
+
</p>
|
96 |
+
</td>
|
97 |
+
</tr>
|
98 |
+
</table>
|
99 |
+
|
100 |
+
## Demo
|
101 |
+
Click here to try out the Demo of [MiniCPM-V](http://120.92.209.146:80).
|