AssertionError: no app_id collisions expected` when scheduling JobGroup with multiple executables (Local Executor)

[bzantium]

Describe the bug

When running a pipeline locally (e.g., executor="none") that involves a JobGroup or multiple bundled tasks in a single step, the execution fails with an AssertionError: no app_id collisions expected.

This occurs because the local scheduler in nemo_run (via torchx) does not correctly handle the dryrun_info for multiple executables within a single group. Instead of maintaining unique information for each executable, it appears to overwrite the dry-run information during iteration, causing the underlying scheduler to detect an App ID collision when it attempts to schedule the tasks.

Steps/Code to reproduce bug

1. Define a simple experiment with multiple tasks bundled together.
2. configure the run to use the local executor.
3. Execute the experiment.

Here is a minimal example snippet:

import nemo_run as run

# Define two simple tasks
task1 = run.Script(inline='echo "Hello Task 1"')
task2 = run.Script(inline='echo "Hello Task 2"')

# Create an experiment
with run.Experiment("local_collision_test") as exp:
    # Add tasks as a bundle/group (this creates a JobGroup internally)
    exp.add([task1, task2], name="my_job_group")

    # Run locally
    # This triggers the AssertionError
    exp.run()

Expected behavior

The local executor should be able to accept a list of tasks (a JobGroup), generate unique App IDs for each task, and execute them sequentially or in parallel without crashing due to ID collisions.

Additional context

Traceback:

  File "/.../site-packages/nemo_run/run/torchx_backend/schedulers/local.py", line 106, in schedule
    app_id = super().schedule(dryrun_info=dryrun_info)
  File "/.../site-packages/torchx/schedulers/local_scheduler.py", line 791, in schedule
    app_id not in self._apps
AssertionError: no app_id collisions expected since uuid4 suffix is used