Skip to content

Various fixes regarding special tokens (embeddings & anti-prompt)#1112

Merged
martindevans merged 3 commits intoSciSharp:masterfrom
Lyrcaxis:special-fixes
Feb 26, 2025
Merged

Various fixes regarding special tokens (embeddings & anti-prompt)#1112
martindevans merged 3 commits intoSciSharp:masterfrom
Lyrcaxis:special-fixes

Conversation

@Lyrcaxis
Copy link
Copy Markdown
Contributor

Fixed some issues with the following:

  • Fixed special tokens parsing as plain text in LLamaEmbedder

  • Fixed anti-prompt passing as plain text in InteractiveExecutor

  • BatchedExecutorBoolQ example now tokenizes special tokens

I wonder if at this point it's worth making the default Tokenize() parameter be special = true instead,
.. because every single one of the library's calls seems to be needing this to be set as true:
image

Although if you think it's safest for users if accidental special are parsed as plain text I can understand that too.

@martindevans martindevans merged commit 2d1f639 into SciSharp:master Feb 26, 2025
6 checks passed
@Lyrcaxis Lyrcaxis deleted the special-fixes branch February 26, 2025 17:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants