Jérémie C. Wenger

That is in fact a feature of beam search, see [this discussion](https://github.com/minimaxir/gpt-2-simple/issues/51), [this implementation](https://github.com/nshepperd/gpt-2/commit/87fe3d7c50de78e31a0578eea7a4b34d6e8217fe) and [this paper](https://arxiv.org/abs/1904.09751)! Temperature-based random sampling and/or top_p (nucleus) sampling are in my experience always preferable...

My pleasure! From the network's perspective meaningful is not particularly relevant. If it's a character-level or bpe model this repetition of characters over and over might still be the one...

Same here, even when following @markriedl's steps, pip still searches for hours. I know these days it would be more straightforward to just use the model at goose.ai, but I...

Thanks! It works again.

Here is a [Colab Notebook](https://drive.google.com/file/d/1WkTEDaaYzP5z3S3lQB2ADoM1cYq7uQlJ/view?usp=sharing) with instructions on how to train it easily directly inside your Google Drive.

@wincentbalin you're right, I updated it to reflect the current state of affairs, it should work now.

Much better late than never @shiffman, thanks so much for your answer! I learnt so much from you, it's a joy to contribute even in such a modest fashion. I...

Hi, yes, here it is. It's the Gutenberg editor, with a Kadence accordion, and inside a Coblocks Masonry. I can paste whatever I have in the clipboard into, say, the...

Hi, Thanks so much for this! I'm afraid I don't see any changes yet. I updated the plugin (1.8.12) yesterday and I cleared the cache, but I do get the...

Hey, Thanks so much for looking into this! To be honest I don't know if that's caused by the accordion scripts or something else in the theme, and I don't...