search

NVIDIA / nim-anywhere

Accelerate your Gen AI with NVIDIA NIM and NVIDIA AI Workbench
https://www.nvidia.com/en-us/ai/
Apache License 2.0
111 stars 64 forks source link

Updated variable names for embedder and reranker config. #63

Open rmkraus opened 1 week ago

rmkraus commented 1 week ago

closes #57

freemansoft commented 1 week ago

Did this change go the right direction. The variables are now named after the model name. I'm assuming to match something in other code. It seems like the old variable names were valid and the code should have been changed to use them. Now it will be weird if someone tries to use a different reranker or embedding model because the variable names include a different model in the name.

rmkraus commented 1 week ago

That's a fair question, and I actually totally agree that the best solution is to retain the old variable names.

Ultimately, what I did was update the existing variables to have the names that the containers were looking for on startup because it feels like a lower risk because it is fewer lines of change.

We will also be migrating to workbench's new Docker Compose support soon as it will be much more resilient: https://docs.nvidia.com/ai-workbench/user-guide/latest/projects/compose.html