Skip to content

Nemotron 3 Super DGX Spark Deployment Guide broken #132

@kristianpaul

Description

@kristianpaul

Describe the bug

Nemotron 3 Super setup steps for the DGX Spark Setup are broken [1]

Steps/Code to reproduce bug

Follow the steps from the guide (https://docs.nvidia.com/nemotron/nightly/usage-cookbook/Nemotron-3-Super/SparkDeploymentGuide/README.html#vllm) to get vLLM running from docker image and available on port 8000

Expected behavior

Resulting openai compatible endpoint isnt working (tested from latest version of Pi and Opencode).

Additional context

Might be related with a vllm reasoning config issue upstream but despite that the steps are not functioning.

Sometimes you can see how the agent leaks or messages in the conversation.

[1] https://docs.nvidia.com/nemotron/nightly/usage-cookbook/Nemotron-3-Super/SparkDeploymentGuide/README.html#

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions