Jinseo Kim

@willard-yuan The goal of this project is to prove that NeuralHash cannot guarentee anonymity of image, so I used final form(including LSH) of NeuralHash because that is what Apple accesses.

This happens because `BasicTokenizer`, which is [used as](https://github.com/huggingface/transformers/blob/main/src/transformers/models/clip/tokenization_clip.py#L159) fallback text fix function, [strips accents if `do_lower_case=True`](https://github.com/huggingface/transformers/blob/main/src/transformers/models/bert/tokenization_bert.py#L174-L176). We may fix this by explicitly set `strip_accents` to `False`; [ViT/L-14 tokenizer](https://huggingface.co/openai/clip-vit-large-patch14/raw/main/tokenizer.json) includes vocabs...

Hello! In short: the conversion scripts exists, but they are pure spaghetti and unpolished, so I don't want to publish them. (If you need anyway, I can email you) Recently...

Hello, First of all, thank you for your idea! The notification email was bounced on my inbox so I couldn't reply quickly... 😓 I agree that we can do better...

Thank you for your contribution! Some thoughts: - It seems these patches have largely three points: - Converting the original SD checkpoint to the model which is compatible to this...