Skip to content

fix: return in-scope-filtered agent configs from load_and_validate_server_instance_configs - #1410

Merged
ananthsub merged 1 commit into
NVIDIA-NeMo:mainfrom
ananthsub:fix/train-data-in-scope-return
May 29, 2026
Merged

ananthsub merged 1 commit into
NVIDIA-NeMo:mainfrom
ananthsub:fix/train-data-in-scope-return

Conversation

@ananthsub

Copy link
Copy Markdown
Contributor

load_and_validate_server_instance_configs() builds an agent_configs_with_in_scope_datasets list that drops agents whose datasets are entirely out of scope for the current mode and restricts mixed agents to their in-scope datasets, but the function then returned the unfiltered agent_configs_with_data.

Downstream consumers (load_datasets, validate_samples_and_aggregate_metrics, collate_samples) therefore saw every agent that had any dataset and, for agents whose datasets were all out of scope, never had their data filtered at all.

Fix: return agent_configs_with_in_scope_datasets and fix the misleading count in the accompanying printout

@ananthsub
ananthsub force-pushed the fix/train-data-in-scope-return branch from 2d8c4b8 to a1a7edc Compare May 28, 2026 22:50
@copy-pr-bot

copy-pr-bot Bot commented May 28, 2026

Copy link
Copy Markdown

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.

@ananthsub
ananthsub enabled auto-merge (squash) May 28, 2026 22:50
…ad_and_validate_server_instance_configs

load_and_validate_server_instance_configs() builds an
agent_configs_with_in_scope_datasets list that drops agents whose datasets
are entirely out of scope for the current mode and restricts mixed agents
to their in-scope datasets, but the function then returned the unfiltered
agent_configs_with_data. Downstream consumers (load_datasets,
validate_samples_and_aggregate_metrics, collate_samples) therefore saw
every agent that had any dataset and, for agents whose datasets were all
out of scope, never had their data filtered at all.

Return agent_configs_with_in_scope_datasets and fix the misleading count
in the accompanying printout. Add a regression test that constructs three
in-memory agent configs (in-scope-only, out-of-scope-only, mixed) in
example_validation mode and asserts the out-of-scope-only agent is
dropped while the mixed agent is restricted to its in-scope datasets.

Signed-off-by: Ananth Subramaniam <ansubramania@nvidia.com>
@ananthsub
ananthsub force-pushed the fix/train-data-in-scope-return branch from a1a7edc to be2d67e Compare May 29, 2026 15:52
@ananthsub
ananthsub merged commit 4b3a783 into NVIDIA-NeMo:main May 29, 2026
9 of 10 checks passed
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.

2 participants