meta-llama

/

Llama-3.1-405B-Instruct-FP8

like 186

Follow

Meta Llama 32.2k

Hey, not sure if this matters given that the code path defaults to not quantizing lm head anyways afaict, but "lm_head" is missing from the quantization config for this model but is present for the base 405b fp8 model. Thanks!