The CLIP model (Contrastive Language Image Pre-training) by OpenAI revolutionzed text to image generation. Implementation adapted from https://github.com/moein-shariatnia/OpenAI-CLIP.
Here we implement a quick version of and test it by prompting it with the phrase "cool as hell" :p
The structure of the notebook is supposed to mirror a pytorch module with scripts:
- config.py
- utils.py
- encoders.py
- clip.py
- dataset.py
- dataladers.py
- train.py
- inference.py