fp16 version of the model
#6
by
Light4Bear
- opened
Is there any benefit using fp32? I think the original Llama from meta is already in fp16.
It will run in fp16 in transformers, but it would have been better to have fp16 uploaded so it would be smaller