Closed zkyseu closed 1 year ago
You can simply adjust the top_p and/or top_k and sample multiple times. Please see nucleus sampling widely used in language modeling.
On Wed, Jun 14, 2023 at 5:45 AM zkyseu @.***> wrote:
Hi, thanks for your wonderful work! I have a question that whether pix2seq can generate different inference results by setting random seed? If it can, how can i modify the code?
— Reply to this email directly, view it on GitHub https://github.com/google-research/pix2seq/issues/39, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAKERUK2LI773XAGKIQRGH3XLGXAJANCNFSM6AAAAAAZGJ4GWA . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Thanks!
Hi, thanks for your wonderful work! I have a question that whether pix2seq can generate different inference results by setting random seed? If it can, how can i modify the code?