Can we get the fp16 variant please?
#2
by
natalie5
- opened
I really like the extra precision.
i'm sorry, but we only created an FP8 variant due to resources. the finetune was made from 8-bit, so this model is quantization-native.
i'm sorry, but we only created an FP8 variant due to resources. the finetune was made from 8-bit, so this model is quantization-native.
Oh that's fine, thanks for this model!
natalie5
changed discussion status to
closed
you're welcome!
Could you convert it to GGUF? I'd like to download a 4-bit quant of it to try
sure!