I get an OSError while trying to run the paratope-prediction.ipynb.
OSError: We couldn't connect to 'https://huggingface.co' to load this model, couldn't find it in the cached files and it looks like antiberta-master is not the path to a directory containing a {configuration_file} file.
Checkout your internet connection or see how to run the library in offline mode at 'https://huggingface.co/docs/transformers/installation#offline-mode'.

Can you help me and tell me what I am doing wrong? I have the internet connection. Perhaps MODEL_DIR must be adjusted with the value "antiberta-master"?

Many thanks in advance!

Hi, when I try to load the tokenizer I get following error:
HTTPError: 401 Client Error: Unauthorized for url: https://huggingface.co/antibody-tokenizer/resolve/main/vocab.json

I was wondering if the tokenizer has been removed?

Hey team, great work on this project. I noticed that the pre-training data snippets were removed from the repository back in April. Are they available as a HF dataset? Or available someplace else (besides the git history)?

Hello, I have used the model weights and tokenizer that you uploaded on Hugging Face, but I encountered some diffcults when actually using it. I find the output of tokenizer is weird.Could you tell me Is it necessary to insert a space between every character in the input sequence? and is the use of H and L (on the start of sequence) necessary? thanks!
tokenizer = RoFormerTokenizer.from_pretrained("trans_model/antiberta2") model = RoFormerForMaskedLM.from_pretrained("trans_model/antiberta2") seqs = ["ḢACCKLMNDDDKKLL", "ḶALYYNNMACD"] outs = tokenizer( seqs, return_tensors="pt", padding=True, truncation=True, max_length=1024, )
`

outs
{'input_ids': tensor([[ 1, 26, 4, 2],
[ 1, 27, 4, 2]]), 'token_type_ids': tensor([[0, 0, 0, 0],
[0, 0, 0, 0]]), 'attention_mask': tensor([[1, 1, 1, 1],
[1, 1, 1, 1]])}
seqs = ["Ḣ A C C K L M N D D D K K L L", "Ḷ A L Y Y N N M A C D"]
out
{'input_ids': tensor([[ 1, 26, 5, 6, 6, 13, 14, 15, 16, 7, 7, 7, 13, 13, 14, 14, 2],
[ 1, 27, 5, 14, 24, 24, 16, 16, 15, 5, 6, 7, 2, 0, 0, 0, 0]]), 'token_type_ids': tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]), 'attention_mask': tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0]])}
`

Greetings,

I am trying to fine tune your model in my own dataset. But when I instantiate the tokenizer and model through huggingface,
HTTPError: 401 Client Error: Unauthorized for url: https://huggingface.co/antibody-tokenizer/resolve/main/vocab.json. This error occurs which I think is because this is a private model not available in hugginface?

Can you give me access to the tokenizer and model?

Thankyou,
Palistha

Hi all, congratulations on the Cell Patterns paper. I am keen to give antiberta a try.

On the notebook, where it says "Inference - how to go from sequence to predicted sequence" at the end, does that give a paratope string prediction similar to the Figure 4 sequences with dots in the paper?

I can't find the datasets file which prevents me from converting the data to dateset.I would appreciate it if you could provide the datesets.