"___BEGIN__" in output of make_sentence_with_start() about markovify HOT 4 CLOSED

Strange, and thanks for flagging! Could you share a minimal script and corpus text that reproduces the problem?

from markovify.

I updated the script above. Here's a sample corpus text: https://files.quietmint.ml/corpus.txt

from markovify.

Thanks for flagging this, @quietmint! Looks like it was a bug here: c41af53#diff-6dd95c17c1c64b93c71d240c9b497656

Fix is in 0.7.1, now available via pip. Let me know if it doesn't work for you.

from markovify.

Well that fixed the issue with ___BEGIN__ but now I get None as the output sometimes. This should never happen because there are definitely countless sentences starting with my requested word, and I've set the flag test_output=False to ensure any sentence is acceptable (regardless of overlap percentage). On the model, I've also set the flag retain_original=False (which should imply test_output=False, but regardless I've set it explicitly...)

Does max_words only consider original sentences of the specified length? I thought it would consider any sentence length but stop producing text after X words?

import markovify

text = open('corpus.txt').read()
model = markovify.Text(text, state_size=3, retain_original=False)
print(model.make_sentence_with_start('Who', max_words=7, test_output=False))

from markovify.