GPU requirements

#14
by YorelNation - opened

Hi, thanks for releasing the model open source <3

I currently run the first mistral instruct model on a A100 using vllm.

My question is simple: Can the new model fit on my A100 ?

It depends, if you quantize the model it can fit in 24GB (4bits)

oh really ?
I have an Nvidia tesla P40 how can I run it ?

Hi @DeltaWhiplash
check out this section on how to run the model using different optimization methods: https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1#run-the-model

OMG am blind
Am using a VM with this specs :

  • 8 vCPU
  • 16GB of RAM
  • a p40 24gb passthrough

to run my models am using this inference (https://github.com/c0sogi/llama-api)
Do you have one to recommend (I just need an openai like api)

Same here @DeltaWhiplash

Sign up or log in to comment