Re-quantize and re-upload model

#1
by mtasic85 - opened

@Lyte llama.cpp fixed issue with BOS/EOS tokens, and all models need to be re-quantized and re-uploaded.

Please check https://github.com/ggerganov/llama.cpp/issues/9315

Owner

okay thanks for letting me know, I'll get to it asap!

Owner
β€’
edited Sep 16

@mtasic85 done using latest llama.cpp same as the other repo. feel free to try quantizing them yourself I've included the notebook i use to do this, it's the same one that i used for both RWKV models.

Lyte changed discussion status to closed

Sign up or log in to comment