10000 [BUG] progress_bar printer fails on test due non-instantiated variable · Issue #15064 · NVIDIA-NeMo/NeMo · GitHub
[go: up one dir, main page]

Skip to content

[BUG] progress_bar printer fails on test due non-instantiated variable #15064

@MidAtBest

Description

@MidAtBest

Describe the bug

At the following location,
nemo/lightning/pytorch/callbacks/progress_printer.py(212)on_test_batch_end()

self.total_validation_steps is used instead of self.total_test_steps

Steps/Code to reproduce bug

  • Create a model with the ProgressPrinter callback
  • Setup the megatron strategy and trainer
  • launch trainer.test()

Expected behavior

No failure driven by the non instantiation of self.total_validation_steps. self.total_test_steps should replace this value, hence fixing the issue as this is a test step and not a validation step.

Environment overview (please complete the following information)

  • Environment location: Docker
  • Method of NeMo install: from source (using bionemo-framework)

Environment details

If NVIDIA docker image is used you don't need to specify these.
Otherwise, please provide:

  • OS version: Ubuntu 24.04.3 LTS
  • PyTorch version: 2.8.0
  • Python version: 3.12.3

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions

      0