Skip to content

[Super 3.5 evals] Update Sep 02, 2026 - #2979

Merged
bxyu-nvidia merged 3 commits into
mainfrom
bxyu/super-3.5-evals-202060902
Sep 2, 2026
Merged

bxyu-nvidia merged 3 commits into
mainfrom
bxyu/super-3.5-evals-202060902

Conversation

@bxyu-nvidia

@bxyu-nvidia bxyu-nvidia commented Sep 2, 2026 •

Copy link
Copy Markdown
Contributor

What does this PR do?

  1. Swap to no pty: resources_servers/swebench/app.py, resources_servers/swebench/swebench_patches.py, resources_servers/terminal_bench_2_1/app.py, responses_api_agents/opencode_sandboxed_agent/tests/test_app.py, responses_api_agents/opencode_sandboxed_agent/app.py, responses_api_agents/terminus_2_sandboxed_agent/tests/test_app.py, responses_api_agents/terminus_2_sandboxed_agent/app.py
  2. vLLM router health check counter fix (see feat: add vllm router patch for decode imbalanceΒ #2952): benchmarks/nemotron_3.5_super/build_eval_container.sh, benchmarks/nemotron_3.5_super/build_vllm_router_wheel.sh, benchmarks/nemotron_3.5_super/README.md, benchmarks/nemotron_3.5_super/sbatch_external_vllm.sh
  3. Increase default vllm_model workers: benchmarks/nemotron_3.5_super/policy_model_override.yaml
  4. Silence no-op termination logging from OpenSandbox: nemo_gym/sandbox/providers/opensandbox/provider.py, tests/unit_tests/test_opensandbox_provider.py

Checklist

  • I have read the contributing guidelines.
  • The change is focused; unrelated "drive-by" edits are tracked as separate issues/PRs.
  • Tests added or updated and pass locally, or N/A for docs-only / non-code changes (so CI unit/server checks pass when applicable).
  • Pre-commit checks pass locally (pre-commit run --all-files) (so CI lint/format/copyright pass).
  • All commits have DCO sign-off (git commit -s) (so the DCO check passes).

Signed-off-by: Brian Yu <bxyu@nvidia.com>
Signed-off-by: Brian Yu <bxyu@nvidia.com>
Signed-off-by: Brian Yu <bxyu@nvidia.com>
@bxyu-nvidia
bxyu-nvidia merged commit ba4d686 into main Sep 2, 2026
38 checks passed
@bxyu-nvidia
bxyu-nvidia deleted the bxyu/super-3.5-evals-202060902 branch September 2, 2026 17:44
jkyi-nvidia pushed a commit that referenced this pull request Sep 11, 2026
<!-- Thanks for contributing to NeMo Gym! Please fill out the sections
below. -->

## What does this PR do?

1. Swap to no pty: resources_servers/swebench/app.py,
resources_servers/swebench/swebench_patches.py,
resources_servers/terminal_bench_2_1/app.py,
responses_api_agents/opencode_sandboxed_agent/tests/test_app.py,
responses_api_agents/opencode_sandboxed_agent/app.py,
responses_api_agents/terminus_2_sandboxed_agent/tests/test_app.py,
responses_api_agents/terminus_2_sandboxed_agent/app.py
2. vLLM router health check counter fix (see
#2952):
benchmarks/nemotron_3.5_super/build_eval_container.sh,
benchmarks/nemotron_3.5_super/build_vllm_router_wheel.sh,
benchmarks/nemotron_3.5_super/README.md,
benchmarks/nemotron_3.5_super/sbatch_external_vllm.sh
3. Increase default vllm_model workers:
benchmarks/nemotron_3.5_super/policy_model_override.yaml
4. Silence no-op termination logging from OpenSandbox:
nemo_gym/sandbox/providers/opensandbox/provider.py,
tests/unit_tests/test_opensandbox_provider.py

## Checklist

- [ ] I have read the [contributing
guidelines](https://docs.nvidia.com/nemo/gym/latest/contribute/development-setup).
- [ ] The change is focused; unrelated "drive-by" edits are tracked as
separate issues/PRs.
- [ ] Tests added or updated and pass locally, or N/A for docs-only /
non-code changes (so CI unit/server checks pass when applicable).
- [ ] Pre-commit checks pass locally (`pre-commit run --all-files`) (so
CI lint/format/copyright pass).
- [ ] All commits have DCO sign-off (`git commit -s`) (so the DCO check
passes).

---------

Signed-off-by: Brian Yu <bxyu@nvidia.com>
Signed-off-by: Junkeun Yi <jkyi@nvidia.com>

This branch was previously deployed

1 inactive deployment
public β€” 544008a7 Deployed Sep 2, 2026 by copy-pr-bot[bot] via release / finalize / notify #2445
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.

1 participant