MIssing the Q8 version
#3
by
markstachowski
- opened
I noticed you have the fp16 model added twice and skipped the Q8 version of the model.
Removed the old fp16 model. The newer one has the fixed pre-tokenizer, hence why it was reuploaded to begin with.
Q8 is not skipped, as specified in the model card it is in its own branch.
failspy
changed discussion status to
closed