README.md

---
license: apache-2.0
language:
- en
base_model:
- apple/MobileCLIP2-S4
- apple/MobileCLIP2-S2
pipeline_tag: image-text-to-text
tags:
- MobileCLIP
- MobileCLIP2
- CLIP
- Classification
---

# MobileCLIP2

The following versions of MobileCLIP2 have been converted to run on the Axera NPU using w8a16 quantization. Compatible with Pulsar2 version: 4.2
- MobileCLIP2-S2
- MobileCLIP2-S4

If you want to know how to convert the MobileCLIP2 model into an axmodel that can run on the axera npu board, please read [this link](https://github.com/AXERA-TECH/axera.ml-mobileclip) in detail.

## Support Platform
- AX650

## End-of-board inference time
- MobileCLIP2-S2
| Stage | Time |
  |------|------|
  | image encoder | 19.146 ms  | 
  | text encoder | 5.675 ms  |

-  MobileCLIP2-S4
  | Stage | Time |
  |------|------|
  | image encoder | 65.328 ms  | 
  | text encoder | 12.663 ms  |


## How to use

Download all files from this repository to the device

Run the following command:
```bash
python3 run_axmodel.py -ie ./mobileclip2_s4_image_encoder.axmodel -te ./mobileclip2_s4_text_encoder.axmodel -i ./zebra.jpg -t "a zebra" "a dog" "two zebras"
```

Model input and output examples are as follows:
1. the image you want to input:

    ![](zebra.jpg)
  
3. The description of the text you want to categorize:

    ["a zebra", "a dog", "two zebras"]

4. Model output class confidence scores:

    Label probs: [[6.095444e-02 5.628616e-14 9.390456e-01]]