Tiny CLIP

Introduction

This is a smaller version of CLIP trained for EN only. The training script can be found here. This model is roughly 8 times smaller than CLIP. This was achieved by using a small text model (microsoft/xtremedistil-l6-h256-uncased) and a small vision model (edgenext_small). For a in-depth guide of training CLIP see this blog.

Usage

For now this is the recommended way to use this model

git lfs install 
git clone https://huggingface.co./sachin/tiny_clip
cd tiny_clip

Once you are in the folder you could do the following:

import models
text_encoder, tokenizer, vision_encoder, transform = models.get_model()