Different rms_norm_eps
#6
by
djuna
- opened
Hi,
I noticed that from 14B onward, the value of the base and instruct is different. is this intentional?
hi,
I would like to inquire whether this model supports FP16 deployment with vLLM.
You can run this model with the command vllm serve Qwen/Qwen2.5-7B, but keep in mind it's not set up for chatting.