Markus F
Markus F
Thanks for the sigmoid/np.inf fix. I will adapt this later. **1.** I'll also add that there should be no newlines in the readme. First, I'll test it myself though, but...
I understand that - that would be very helpful - whatever does the job! I don't get the question, could you please reformulate?
I see. The tokenizer takes care of this. If you pass a newline, it will be tokenized in the same way as a space. But we only return character probabilities,...
Hi @eshau @mexus, I finally had the chance to look into this, but I can't really reproduce your error with the embeddings. I tested it using the above script and...
Hi @eshau @mexus, did it work for you? Or is there still anything unclear?
Hi @mexus, thanks for your response! Looking at your `config.json`, I think the issue is setting `"one_sample_per_line": true"`. But this should be set to `false`! Let me know if it...
That's great, happy to hear! Thanks for confirming. Then I'll close the issue.
Hi @ir2718 , thanks for catching that - weird it never occured before! Can you maybe share the text or a minimal example where this happens? I'll then look into...
@ir2718 it would be great if you could share an example so I can look into that - thanks!
Hi @rehandaphedar! Indeed, there is a new version incompatibility. Thanks for catching! You can fix it with the latest `requirements.txt` I provided in the latest push. See https://github.com/segment-any-text/wtpsplit/blob/main/requirements.txt It just...