Add task_count

[hiyuchang]

Description

[Please describe the background, purpose, changes made, and how to test this PR]

Checklist

Please check the following items before code is ready to be reviewed.

• Code has passed all tests
• Docstrings have been added/updated in Google Style
• Documentation has been updated
• Code is ready for review

[gemini-code-assist]

Summary of Changes

Hello @hiyuchang, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request enhances the training process by adding a new metric, task_count, to the trainer. This metric provides valuable insight into the diversity of tasks being handled in each training step, which can aid in monitoring and understanding the training dynamics without altering the core training logic.

Highlights

• New Metric Added: Introduced a new metric named task_count within the train_step function in trinity/trainer/trainer.py.
• Task Uniqueness Tracking: The task_count metric calculates the number of unique tasks processed during a training step by counting distinct task attributes from the eids in the exps object.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request adds a task_count metric during the training step. The logic for calculating the number of unique tasks is correct. However, there's a minor issue with the order of metric updates that could cause the new task_count to be overwritten. I've provided a suggestion to fix this potential bug and slightly improve performance.

[pan-x-c]

/unittest-module-trainer

[github-actions]

Summary

Tests 📝	Passed ✅	Failed ❌	Skipped ⏭️	Other ❓	Flaky 🍂	Duration ⏱️
20	17	1	2	0	0	2.0s

Failed Tests

Failed Tests ❌	Fail Message
❌ tests/trainer/trainer_test.py::TestTrainerLoRA::test_trainer	The test failed in the call phase due to an assertion error

Skipped

Tests	Status
tests/trainer/trainer_test.py::TestMultiModalGRPO::test_trainer	skipped ⏭️
tests/trainer/trainer_test.py::TestMultiModalSFT::test_trainer	skipped ⏭️

Tests

Test Name	Status	Flaky	Duration
tests/trainer/trainer_test.py::TestTrainerCountdown_0_fsdp::test_trainer	✅		137ms
tests/trainer/trainer_test.py::TestTrainerCountdown_1_megatron::test_trainer	✅		323ms
tests/trainer/trainer_test.py::TestStepAheadAsyncRL::test_trainer	✅		61ms
tests/trainer/trainer_test.py::TestTrainerGSM8K_0_fsdp::test_trainer	✅		59ms
tests/trainer/trainer_test.py::TestTrainerGSM8K_1_fsdp2::test_trainer	✅		53ms
tests/trainer/trainer_test.py::TestTrainerGSM8K_2_fsdp::test_trainer	✅		59ms
tests/trainer/trainer_test.py::TestTrainerGSM8K_3_fsdp2::test_trainer	✅		62ms
tests/trainer/trainer_test.py::TestTrainerSFTWarmupGSM8K::test_trainer	✅		104ms
tests/trainer/trainer_test.py::TestTrainerDPO::test_trainer	✅		40ms
tests/trainer/trainer_test.py::TestTrainerSFT::test_trainer	✅		36ms
tests/trainer/trainer_test.py::TestTrainerToolsSFT::test_trainer_tools	✅		36ms
tests/trainer/trainer_test.py::TestFullyAsyncMode_0_fsdp::test_fully_async_mode	✅		87ms
tests/trainer/trainer_test.py::TestFullyAsyncMode_1_fsdp::test_fully_async_mode	✅		84ms
tests/trainer/trainer_test.py::TestFullyAsyncMode_2_megatron::test_fully_async_mode	✅		171ms
tests/trainer/trainer_test.py::TestTrainerCheckpointSave_0_fsdp::test_trainer	✅		103ms
tests/trainer/trainer_test.py::TestTrainerCheckpointSave_1_megatron::test_trainer	✅		347ms
tests/trainer/trainer_test.py::TestTrainerMIX::test_trainer	✅		58ms
tests/trainer/trainer_test.py::TestMultiModalGRPO::test_trainer	⏭️		1ms
tests/trainer/trainer_test.py::TestMultiModalSFT::test_trainer	⏭️		1ms
tests/trainer/trainer_test.py::TestTrainerLoRA::test_trainer	❌		145ms

Github Test Reporter by CTRF 💚

[hiyuchang]

/unittest-module-trainer

[pan-x-c]

/unittest-module-trainer

[github-actions]

Summary

Tests 📝	Passed ✅	Failed ❌	Skipped ⏭️	Other ❓	Flaky 🍂	Duration ⏱️
20	18	0	2	0	0	2.0s

Skipped

Tests	Status
tests/trainer/trainer_test.py::TestMultiModalGRPO::test_trainer	skipped ⏭️
tests/trainer/trainer_test.py::TestMultiModalSFT::test_trainer	skipped ⏭️

Tests

Test Name	Status	Flaky	Duration
tests/trainer/trainer_test.py::TestTrainerCountdown_0_fsdp::test_trainer	✅		148ms
tests/trainer/trainer_test.py::TestTrainerCountdown_1_megatron::test_trainer	✅		322ms
tests/trainer/trainer_test.py::TestStepAheadAsyncRL::test_trainer	✅		68ms
tests/trainer/trainer_test.py::TestTrainerGSM8K_0_fsdp::test_trainer	✅		52ms
tests/trainer/trainer_test.py::TestTrainerGSM8K_1_fsdp2::test_trainer	✅		53ms
tests/trainer/trainer_test.py::TestTrainerGSM8K_2_fsdp::test_trainer	✅		55ms
tests/trainer/trainer_test.py::TestTrainerGSM8K_3_fsdp2::test_trainer	✅		62ms
tests/trainer/trainer_test.py::TestTrainerSFTWarmupGSM8K::test_trainer	✅		99ms
tests/trainer/trainer_test.py::TestTrainerDPO::test_trainer	✅		39ms
tests/trainer/trainer_test.py::TestTrainerSFT::test_trainer	✅		36ms
tests/trainer/trainer_test.py::TestTrainerToolsSFT::test_trainer_tools	✅		36ms
tests/trainer/trainer_test.py::TestFullyAsyncMode_0_fsdp::test_fully_async_mode	✅		81ms
tests/trainer/trainer_test.py::TestFullyAsyncMode_1_fsdp::test_fully_async_mode	✅		82ms
tests/trainer/trainer_test.py::TestFullyAsyncMode_2_megatron::test_fully_async_mode	✅		174ms
tests/trainer/trainer_test.py::TestTrainerCheckpointSave_0_fsdp::test_trainer	✅		107ms
tests/trainer/trainer_test.py::TestTrainerCheckpointSave_1_megatron::test_trainer	✅		355ms
tests/trainer/trainer_test.py::TestTrainerMIX::test_trainer	✅		58ms
tests/trainer/trainer_test.py::TestMultiModalGRPO::test_trainer	⏭️		1ms
tests/trainer/trainer_test.py::TestMultiModalSFT::test_trainer	⏭️		1ms
tests/trainer/trainer_test.py::TestTrainerLoRA::test_trainer	✅		167ms

Github Test Reporter by CTRF 💚