Multi GPU not supported for local models - Rasa CALM - Rasa Community Forum

👋 Introducing the Rasa Playground

Learn more here →

Multi GPU not supported for local models

post by MannavaVivek on May 22, 2024

I am using Ollama as my platform for running local models. I have a 4x16GB GPU cluster that comfortably supports a llama3 model which is around 40GB. When I just run Ollama server and send a request with curl, it loads the model across all 4 GPUs, and everything works. But when with the same config and server, Rasa sends a request, it tries to load the entire model into one GPU for some reason and the server fails and it never receives a response. How do I specify it to use the available multiple GPUs?

506 views

Related topics

Topic Replies Views Activity
Training RASA model in multiple GPU
Rasa Open Source
0 383 Oct 2021
Run NLU training on multiple GPUs
Rasa Open Source
6 2.2k Nov 2022
Training: How to use multiple GPUs for distributed training?
Rasa Open Source
0 404 Apr 2022
How to use Ollama models in Rasa CALM?
Rasa CALM
8 1.3k Sep 2024
Running model on parallel on google cloud
[[Deprecated]0m Rasa X Community Edition
0 262 Aug 2021