fix: reward profiling not require usage

[cmunley1]

(Gym) phtran@eos0065:~/lustre_phtran/Gym$ ng_collect_rollouts \
    +agent_name=verifiers_agent \
    +input_jsonl_fpath=responses_api_agents/verifiers_agent/data/acereason-math-example.jsonl \
    +output_jsonl_fpath=responses_api_agents/verifiers_agent/data/acereason-math-example-rollouts.jsonl \
    +limit=5
Limiting the number of rows to 5
Using `verifiers_agent` for rows that do not already have an agent ref
Repeating rows 1 times (in a pattern of abc to aabbcc)!
Reading rows: 4it [00:00, 12291.00it/s]
Clearing output fpath since `resume_from_cache=False`!
Collecting rollouts: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5/5 [00:50<00:00, 10.01s/it]
Traceback (most recent call last):
  File "/lustre/fsw/coreai_dlalgo_genai/phtran/Gym/.venv/bin/ng_collect_rollouts", line 10, in <module>
    sys.exit(collect_rollouts())
             ^^^^^^^^^^^^^^^^^^
  File "/lustre/fsw/coreai_dlalgo_genai/phtran/Gym/nemo_gym/rollout_collection.py", line 357, in collect_rollouts
    asyncio.run(rch.run_from_config(config))
  File "/home/phtran/.local/share/uv/python/cpython-3.12.12-linux-x86_64-gnu/lib/python3.12/asyncio/runners.py", line 195, in run
    return runner.run(main)
           ^^^^^^^^^^^^^^^^
  File "/home/phtran/.local/share/uv/python/cpython-3.12.12-linux-x86_64-gnu/lib/python3.12/asyncio/runners.py", line 118, in run
    return self._loop.run_until_complete(task)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/phtran/.local/share/uv/python/cpython-3.12.12-linux-x86_64-gnu/lib/python3.12/asyncio/base_events.py", line 691, in run_until_complete
    return future.result()
           ^^^^^^^^^^^^^^^
  File "/lustre/fsw/coreai_dlalgo_genai/phtran/Gym/nemo_gym/rollout_collection.py", line 279, in run_from_config
    group_level_metrics, agent_level_metrics = rp.profile_from_data(rows, results)
                                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/lustre/fsw/coreai_dlalgo_genai/phtran/Gym/nemo_gym/reward_profile.py", line 91, in profile_from_data
    result = result | result["response"].get("usage", dict())
             ~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
TypeError: unsupported operand type(s) for |: 'dict' and 'NoneType'

verifiers agent is crashing due to recent change in ng collect rollouts which requires usage field from agents. this makes the fall back not error.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.