File size: 6,405 Bytes
401c3b8 b0ba6f3 401c3b8 |
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 |
---
license: apple-amlr
library_name: mobileclip
---
# MobileCLIP2: Improving Multi-Modal Reinforced Training
MobileCLIP2 was introduced in [MobileCLIP2: Improving Multi-Modal Reinforced Training](http://arxiv.org/abs/2508.20691) (TMLR August 2025 <mark>Featured</mark>), by Fartash Faghri, Pavan Kumar Anasosalu Vasu, Cem Koc, Vaishaal Shankar, Alexander T Toshev, Oncel Tuzel, Hadi Pouransari.
This repository contains the **CoCa** checkpoint pretrained on DFN-2B dataset and fine-tuned on varying datasets.
### Highlights
* `MobileCLIP2-S4` matches the accuracy of SigLIP-SO400M/14 with 2x fewer parameters and surpasses DFN ViT-L/14 at 2.5x lower latency measured on iPhone12 Pro Max.
* `MobileCLIP-S3/S4` are our new architectures trained on MobileCLIP’s training dataset, DataCompDR-1B (dashed lines).
* Our smallest variant `MobileCLIP-S0` obtains similar zero-shot performance as [OpenAI](https://arxiv.org/abs/2103.00020)'s ViT-B/16 model while being 4.8x faster and 2.8x smaller.
* `MobileCLIP-S2` obtains better avg zero-shot performance than [SigLIP](https://arxiv.org/abs/2303.15343)'s ViT-B/16 model while being 2.3x faster and 2.1x smaller, and trained with 3x less seen samples.
* `MobileCLIP-B (LT)` attains zero-shot ImageNet performance of **77.2%** which is significantly better than recent works like [DFN](https://arxiv.org/abs/2309.17425) and [SigLIP](https://arxiv.org/abs/2303.15343) with similar architectures or even [OpenAI's ViT-L/14@336](https://arxiv.org/abs/2103.00020).
## Checkpoints
| Model | # Seen <BR>Samples (B) | # Params (M) <BR> (img + txt) | Latency (ms) <BR> (img + txt) | IN-1k Zero-Shot <BR> Top-1 Acc. (%) | Avg. Perf. (%) <BR> on 38 datasets |
|:----------------------------------------------------------|:----------------------:|:-----------------------------:|:-----------------------------:|:-----------------------------------:|:----------------------------------:|
| [MobileCLIP2-S0](https://hf.co/apple/MobileCLIP2-S0) | 13 | 11.4 + 42.4 | 1.5 + 1.6 | 71.5 | 59.7 |
| [MobileCLIP2-S2](https://hf.co/apple/MobileCLIP2-S2) | 13 | 35.7 + 63.4 | 3.6 + 3.3 | 77.2 | 64.1 |
| [MobileCLIP2-B](https://hf.co/apple/MobileCLIP2-B) | 13 | 86.3 + 63.4 | 10.4 + 3.3 | 79.4 | 65.8 |
| [MobileCLIP2-S3](https://hf.co/apple/MobileCLIP2-S3) | 13 | 125.1 + 123.6 | 8.0 + 6.6 | 80.7 | 66.8 |
| [MobileCLIP2-L/14](https://hf.co/apple/MobileCLIP2-L-14) | 13 | 304.3 + 123.6 | 57.9 + 6.6 | 81.9 | 67.8 |
| [MobileCLIP2-S4](https://hf.co/apple/MobileCLIP2-S4) | 13 | 321.6 + 123.6 | 19.6 + 6.6 | 81.9 | 67.5 |
| [MobileCLIP-S0](https://hf.co/apple/MobileCLIP-S0) | 13 | 11.4 + 42.4 | 1.5 + 1.6 | 67.8 | 58.1 |
| [MobileCLIP-S1](https://hf.co/apple/MobileCLIP-S1) | 13 | 21.5 + 63.4 | 2.5 + 3.3 | 72.6 | 61.3 |
| [MobileCLIP-S2](https://hf.co/apple/MobileCLIP-S2) | 13 | 35.7 + 63.4 | 3.6 + 3.3 | 74.4 | 63.7 |
| [MobileCLIP-B](https://hf.co/apple/MobileCLIP-B) | 13 | 86.3 + 63.4 | 10.4 + 3.3 | 76.8 | 65.2 |
| [MobileCLIP-B (LT)](https://hf.co/apple/MobileCLIP-B-LT) | 36 | 86.3 + 63.4 | 10.4 + 3.3 | 77.2 | 65.8 |
| [MobileCLIP-S3](https://hf.co/apple/MobileCLIP-S3) | 13 | 125.1 + 123.6 | 8.0 + 6.6 | 78.3 | 66.3 |
| [MobileCLIP-L/14](https://hf.co/apple/MobileCLIP-L-14) | 13 | 304.3 + 123.6 | 57.9 + 6.6 | 79.5 | 66.9 |
| [MobileCLIP-S4](https://hf.co/apple/MobileCLIP-S4) | 13 | 321.6 + 123.6 | 19.6 + 6.6 | 79.4 | 68.1 |
## How to Use
First, download the desired checkpoint visiting one of the links in the table above, then click the `Files and versions` tab, and download the PyTorch checkpoint.
For programmatic downloading, if you have `huggingface_hub` installed, you can also run:
```
hf download apple/mobileclip2_coca_dfn2b_s13b_<finetune-dataset>_context<length>
```
For models length with context lengths 128/256, copy `config.json` to `src/open_clip/model_configs/coca_ViT-L-14-context$len.json` and change the model name in below example to `coca_ViT-L-14-context$len`.
```py
import torch
import open_clip
from PIL import Image
model, _, preprocess = open_clip.create_model_and_transforms('coca_ViT-L-14', pretrained='/path/to/mobileclip2_coca.pt')
model.eval()
image = preprocess(Image.open("docs/fig_accuracy_latency.png").convert('RGB')).unsqueeze(0)
with torch.no_grad(), torch.cuda.amp.autocast():
syn_text = model.generate(
image,
generation_type="top_p",
top_p=0.9,
fixed_output_length=True
)[0]
syn_text = open_clip.decode(syn_text).split("<end_of_text>")[0].split("<start_of_text>")[-1].split(".")[0].rstrip()
print("Caption:", syn_text)
```
|