refactor(training): remove unused learning_rate_dict multitask handling

[OutisLi]

Remove the unused learning_rate_dict configuration option that allowed per-task learning rate settings in multitask training. This simplifies the code by using a single learning_rate configuration for all tasks.

Changes:

• Remove learning_rate_dict branch in loss initialization
• Remove per-task lr_schedule dictionary creation
• Remove isinstance(dict) check in training loop
• Unify single-task and multi-task code paths

Both PyTorch and Paddle backends are updated consistently.

Summary by CodeRabbit

• Refactor
  • Streamlined learning rate scheduling in multi-task training scenarios to ensure consistent initialization and computation of learning rates across all models.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between a3db25a and 1269ed8.

📒 Files selected for processing (2)

• deepmd/pd/train/training.py
• deepmd/pt/train/training.py

---

📝 Walkthrough

Walkthrough

This PR simplifies multi-task learning rate handling in both PD and PT training modules by removing per-model learning rate dictionaries and conditional branching. Learning rate schedules are now consistently derived from a single config["learning_rate"] source with unified parameter extraction.

Changes

Cohort / File(s)	Summary
Learning Rate Schedule Simplification deepmd/pd/train/training.py, deepmd/pt/train/training.py	Removed per-model learning rate dictionaries and conditional branching logic for multi-task training. Learning rate schedules now always constructed from single config["learning_rate"] source; loss computation uses unified config["learning_rate"]["start_lr"] extraction; training step directly calls self.lr_schedule.value(_step_id) for current learning rate.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• chore: refactor training loop #4435: Refactors multi-task model selection and training step signature in the same files, related to this PR's multi-task learning rate simplification.

Suggested labels

Python

Suggested reviewers

• njzjz

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 33.33% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and specifically describes the main change: removing unused learning_rate_dict multitask handling, which aligns with the PR's core objective of simplifying training code by eliminating per-task learning rate configuration.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Copilot]

Pull request overview

Removes the unused learning_rate_dict multi-task configuration path from the PyTorch and Paddle training backends, standardizing multi-task training to use a single global learning_rate schedule.

Changes:

• Simplified multi-task loss initialization to always use learning_rate.start_lr.
• Removed per-task learning-rate schedule construction and dict-based handling.
• Unified the training-step learning-rate lookup to a single BaseLR schedule for both single- and multi-task modes.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
deepmd/pt/train/training.py	Drops learning_rate_dict branches and dict-based LR schedule usage in PT training/loss initialization and step loop.
deepmd/pd/train/training.py	Mirrors the PT cleanup in the Paddle backend to consistently use a single LR schedule.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.13%. Comparing base (f0a966b) to head (1269ed8).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5278      +/-   ##
==========================================
+ Coverage   81.94%   82.13%   +0.19%     
==========================================
  Files         750      753       +3     
  Lines       75456    75808     +352     
  Branches     3648     3649       +1     
==========================================
+ Hits        61831    62267     +436     
+ Misses      12457    12374      -83     
+ Partials     1168     1167       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.