Missing BERT code?

[codyseally]

I was trying to run your code with some new ideas, but it breaks at this line:

text_embeddings = np.load("rumor_detection_acl2017/output_bert.npy")

How do you generate the output_bert.npy file ? I can't seem to find the relevant code in the repo.

[j0k3w0n]

I found same issue. I think it's numpy data saved from training done in section 3 with RoBERTa. Did you find any kind of fix??

[codyseally]

I found same issue. I think it's numpy data saved from training done in section 3 with RoBERTa. Did you find any kind of fix??

Sadly no. They don't include the BERT code that does the embedding. You can do it yourself though. Look for the library huggingface bert. It has everything plug and play.

This: https://github.com/huggingface/transformers

With it you should be able to parse the tweets and encode them. Best of luck.

[j0k3w0n]

Thank you for the tips. I'll try it. Although it would be great if they could provide the code in order to reproduce the results.