fix(dpmodel): preserve model dictionaries during serialization

[njzjz]

Summary

• preserve the caller-owned model dictionary while save_dp_model() rewrites variables into YAML or HDF5 references;
• copy only the nested dict and list containers that traversal mutates, so large arrays are not duplicated;
• retain support for non-copyable variable objects such as h5py.Dataset;
• add regressions for nested YAML/native serialization and an HDF5-backed variable.

Why existing tests missed this

The existing save/load tests pass deepcopy(self.model_dict) into save_dp_model() and then validate only the serialized round trip. That disposable copy hides mutations of the object supplied by a real caller, so the shallow-copy bug was never observed.

The new tests pass a caller-owned nested structure directly, assert that its containers and variables remain unchanged, and cover both serialization implementations. The HDF5 dataset case also prevents a memory-heavy full deepcopy from becoming the fix.

Validation

• targeted serialization tests and existing save/load tests: 6 passed;
• ruff format .;
• ruff check .;
• git diff --check.

Closes #5639.

Coding agent: Codex
Codex version: codex-cli 0.144.1
Model: gpt-5.6-sol
Reasoning effort: xhigh

Summary by CodeRabbit

• Bug Fixes
  • Saving models no longer alters the original nested data structures.
  • Preserves shared and non-copyable objects, including HDF5-backed datasets, during serialization.
  • Ensures saved models retain expected data when loaded across supported formats.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 75f041f5-4316-445a-a296-97b013364bc8

📥 Commits

Reviewing files that changed from the base of the PR and between 7d5ad38 and 1daea00.

📒 Files selected for processing (2)

• deepmd/dpmodel/utils/serialization.py
• source/tests/common/dpmodel/test_serialization.py

---

📝 Walkthrough

Walkthrough

save_dp_model now traverses an independent dict/list container tree while preserving variable references. Tests cover unchanged nested inputs for YAML/native outputs and HDF5 dataset handling.

Changes

Serialization mutation isolation

Layer / File(s)	Summary
Container-only copy in save_dp_model deepmd/dpmodel/utils/serialization.py	Adds recursive copying for mutable containers and uses it before model traversal, leaving variable objects shared.
Mutation and HDF5 regression coverage source/tests/common/dpmodel/test_serialization.py	Verifies input preservation across YAML and native outputs, including nested arrays and HDF5 datasets.

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested reviewers: copilot

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly describes the main change: preventing model dictionary mutation during serialization.
Linked Issues check	✅ Passed	The code and tests address #5639 by avoiding container mutation during save_dp_model and adding regression coverage for nested structures.
Out of Scope Changes check	✅ Passed	The changes stay focused on serialization behavior and regression tests, with no obvious unrelated scope creep.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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.}

[codecov]

Codecov Report

❌ Patch coverage is 88.88889% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 79.58%. Comparing base (7d5ad38) to head (1daea00).

Files with missing lines	Patch %	Lines
deepmd/dpmodel/utils/serialization.py	88.88%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5788      +/-   ##
==========================================
- Coverage   79.85%   79.58%   -0.27%     
==========================================
  Files        1022     1022              
  Lines      117351   117358       +7     
  Branches     4313     4311       -2     
==========================================
- Hits        93706    93401     -305     
- Misses      22101    22409     +308     
- Partials     1544     1548       +4     

☔ View full report in Codecov by Harness.
📢 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.

[wanghan-iapcm]

Correct and minimal. save_dp_model did a shallow model_dict.copy(), but traverse_model_dict reassigns nested dict/list entries in place, so serialization mutated the caller's dict (arrays replaced by YAML/HDF5 references). _copy_model_containers rebuilds only the container tree and shares the variable objects, which is safe because the save callbacks read variables (data=x, x.dtype, x.tolist()) and create new @is_variable dicts rather than mutating inputs. The regression tests genuinely fail pre-fix -- passing a caller-owned dict directly and asserting equality/identity -- and the h5py.Dataset test additionally prevents a naive full deepcopy from being the fix, since Datasets aren't deep-copyable. Good catch on why the existing tests missed it (they fed in a deepcopy). LGTM.