@indutny if you end up using it in gpt4all I suggest tuning the parameters base on the demo of the model on the official source. Default gpt4all is too "noisy" so you'll get worse results (more likely to pick tokens it doesn't think are as likely) for code at least. Dunno if that's somehow better for speech.
@indutny if you end up using it in gpt4all I suggest tuning the parameters base on the demo of the model on the official source. Default gpt4all is too "noisy" so you'll get worse results (more likely to pick tokens it doesn't think are as likely) for code at least. Dunno if that's somehow better for speech.