Skip to content

Conversation

@johnnygreco
Copy link
Contributor

This PR updates the default Nemotron 3 Nano inference parameters to what the team tuned it to: temp=1, top_p=1.

I also updated how we store default model settings. The updates make it so

  • All the default settings are in one file (constants.py).
  • We can easily override default settings when necessary.

sanity check comparing old vs new default config file:

➜ diff  ~/.data-designer/model_configs.yaml ~/Desktop/model_configs.yaml
9,10c9,10
<     temperature: 1.0
<     top_p: 1.0
---
>     temperature: 0.85
>     top_p: 0.95

Comment on lines +311 to +317
"text": {"model": "nvidia/nemotron-3-nano-30b-a3b", "inference_parameters": {"temperature": 1.0, "top_p": 1.0}},
"reasoning": {"model": "openai/gpt-oss-20b", "inference_parameters": DEFAULT_REASONING_INFERENCE_PARAMS},
"vision": {"model": "nvidia/nemotron-nano-12b-v2-vl", "inference_parameters": DEFAULT_VISION_INFERENCE_PARAMS},
"embedding": {
"model": "nvidia/llama-3.2-nv-embedqa-1b-v2",
"inference_parameters": DEFAULT_EMBEDDING_INFERENCE_PARAMS | {"extra_body": {"input_type": "query"}},
},
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

main update is to keep the default models next to their default settings

nabinchha
nabinchha previously approved these changes Dec 16, 2025
Copy link
Contributor

@nabinchha nabinchha left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🚢

@johnnygreco johnnygreco merged commit 60c1aed into main Dec 17, 2025
13 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants