Describe the bug

silent training hangs using math with judge or genrm using local vllm model

Steps/Code to reproduce bug

Please list minimal steps or code snippet for us to be able to reproduce the bug.

A helpful guide on on how to craft a minimal bug report http://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports.

Expected behavior

A clear and concise description of what you expected to happen.

Configs
NeMo Gym (e.g. via ng_dump_config) or RL training framework config files.

Environment details

Otherwise, please provide:

• OS version
• Python version
• uv pip list output

Additional context

Add any other context about the problem here.
Example: GPU model