Replies: 2 comments
-
I just added in the pre-tokenizer quickly and converted. But on running, you get errors related to incorrect tensor shapes. Namely, the embedding is From the look of it, only Gemma and Gemma2 currently do something similar. |
Beta Was this translation helpful? Give feedback.
0 replies
-
Looks like work is underway. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Per the ticket template for feature requests, how does the community feel about adding support for Mistral-Nemo-Instruct?
On its face, it looks like there's a need to add support for a pre-tokenizer type called
mistral-bpe
.Beta Was this translation helpful? Give feedback.
All reactions