Hugging Face's logo

visheratin
/
mexma-siglip2

Zero-Shot Image Classification
ONNX
Safetensors
siglip2
clip
mexma
custom_code
Eval Results
Model card Files Community
3
mexma-siglip2
File size: 2,516 Bytes 
87d91da
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
 
---
license: mit
language:
- ar
- kn
- ar
- ka
- af
- kk
- am
- km
- ar
- ky
- ar
- ko
- as
- lo
- az
- ml
- az
- mr
- be
- mk
- bn
- my
- bs
- nl
- bg
- ca
- 'no'
- cs
- ne
- ku
- pl
- cy
- pt
- da
- ro
- de
- ru
- el
- sa
- en
- si
- eo
- sk
- et
- sl
- eu
- sd
- fi
- so
- fr
- es
- gd
- sr
- ga
- su
- gl
- sv
- gu
- sw
- ha
- ta
- he
- te
- hi
- th
- hr
- tr
- hu
- ug
- hy
- uk
- id
- ur
- is
- vi
- it
- xh
- jv
- zh
- ja
pipeline_tag: zero-shot-image-classification
tags:
- siglip2
- clip
- mexma
model-index:
  - name: mexma-siglip2
    results:
      - task:
          type: zero-shot retrieval
        dataset:
          name: Crossmodal-3600
          type: Crossmodal-3600
        metrics:
          - name: Image retrieval R@1
            type: Image retrieval R@1
            value: 62.54%
          - name: Text retrieval R@1
            type: Text retrieval R@1
            value: 59.99%
---

## Model Summary

MEXMA-SigLIP2 is a model that combines the [MEXMA](https://huggingface.co./facebook/MEXMA) multilingual text encoder and an image encoder from the 
[SigLIP2](https://huggingface.co./google/siglip2-so400m-patch16-512/) model. This allows us to get a high-performance CLIP model for 80 languages. 
MEXMA-SigLIP2 sets new state-of-the-art on the [Crossmodal-3600](https://google.github.io/crossmodal-3600/) dataset with 62.54% R@1 for image retrieval and 
59.99% R@1 for text retrieval.


## How to use

```
from transformers import AutoModel, AutoTokenizer, AutoImageProcessor
from PIL import Image
import requests
import torch

model = AutoModel.from_pretrained("visheratin/mexma-siglip2", torch_dtype=torch.bfloat16, trust_remote_code=True, optimized=True).to("cuda")
tokenizer = AutoTokenizer.from_pretrained("visheratin/mexma-siglip2")
processor = AutoImageProcessor.from_pretrained("visheratin/mexma-siglip2")

img = Image.open(requests.get("https://static.independent.co.uk/s3fs-public/thumbnails/image/2014/03/25/12/eiffel.jpg", stream=True).raw)
img = processor(images=img, return_tensors="pt")["pixel_values"]
img = img.to(torch.bfloat16).to("cuda")
with torch.inference_mode():
    text = tokenizer(["кошка", "a dog", "एफिल टॉवर"], return_tensors="pt", padding=True).to("cuda")
    image_logits, text_logits = model.get_logits(text["input_ids"], text["attention_mask"], img)
    probs = image_logits.softmax(dim=-1)
    print(probs)
```

## Acknowledgements

I thank [ML Collective](https://mlcollective.org/) for providing compute resources to train the model.