Phi-Marketing
#336
by
mangojesus
- opened
ahhhh i dont see tokenizers in the file... so, if there is no tokenizer in the model files, then it cant be quantized?
yeah, without any tokenizer, llama.cpp wouldn't know how to convert the text iput/output to/from tokens that the model needs. Probably it's using the same tokenizer as the base model, but llama.cpp needs the tokenizer.
mradermacher
changed discussion status to
closed
sorry for throwing the three models up face palm i caught it after they were up. thanks for clarifying. i will reach out to the model devs and be like, yo.... uppit puppet with the tokenizers. sheesh.
no harm done, don't worry