fix(pt,pd): remove redundant tensor handling to eliminate tensor construction warnings

[Copilot]

This PR fixes deprecation warnings that occur when torch.tensor() or paddle.to_tensor() is called on existing tensor objects:

PyTorch warning:

UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).

PaddlePaddle warning:

UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach(), rather than paddle.to_tensor(sourceTensor).

Root Cause

The warnings were being triggered in multiple locations:

1. PyTorch: Test cases were passing tensor objects directly to ASE calculators, which internally convert them using torch.tensor()
2. PaddlePaddle: Similar issues in eval_model function and to_paddle_tensor utility, plus a TypeError where tensor.to() method was incorrectly using place= instead of device=

Solution

For PyTorch:

• Modified test cases to convert tensor inputs to numpy arrays before passing to ASE calculators
• Removed redundant tensor handling in to_torch_tensor utility function since the non-numpy check already handles tensors by returning them as-is

For PaddlePaddle:

• Added proper type checking in eval_model function to handle existing tensors with clone().detach()
• Removed redundant tensor handling in to_paddle_tensor utility function, applying the same optimization as PyTorch
• Fixed TypeError by changing place= to device= in all tensor.to() method calls (PaddlePaddle's tensor .to() method expects device= parameter, while paddle.to_tensor() correctly uses place=)

Changes Made

1. source/tests/pt/test_calculator.py: Fixed TestCalculator and TestCalculatorWithFparamAparam to convert PyTorch tensors to numpy arrays before passing to ASE calculator
2. deepmd/pt/utils/utils.py: Removed redundant tensor-specific handling in to_torch_tensor function
3. source/tests/pd/common.py: Updated eval_model function with type checking for PaddlePaddle tensors and fixed tensor.to() method calls to use device= instead of place=
4. deepmd/pd/utils/utils.py: Removed redundant tensor-specific handling in to_paddle_tensor function for consistency with PyTorch

Both utility functions now use a simplified approach where the if not isinstance(xx, np.ndarray): return xx check handles all non-numpy inputs (including tensors) by returning them unchanged, eliminating the need for separate tensor-specific code paths.

This change is backward compatible and maintains the same functionality while eliminating both deprecation warnings and TypeErrors, improving code consistency between PyTorch and PaddlePaddle backends.

Fixes #3790.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.29%. Comparing base (6dac4f9) to head (a42db43).
⚠️ Report is 3 commits behind head on devel.

Additional details and impacted files

@@           Coverage Diff           @@
##            devel    #4907   +/-   ##
=======================================
  Coverage   84.29%   84.29%           
=======================================
  Files         702      702           
  Lines       68665    68664    -1     
  Branches     3573     3572    -1     
=======================================
+ Hits        57882    57883    +1     
+ Misses       9643     9642    -1     
+ Partials     1140     1139    -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.

[njzjz]

@copilot There are still warnings:

source/tests/pt/test_calculator.py::TestCalculatorWithFparamAparam::test_calculator
  /opt/hostedtoolcache/Python/3.9.23/x64/lib/python3.9/site-packages/torch/utils/_device.py:103: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.detach().clone() or sourceTensor.detach().clone().requires_grad_(True), rather than torch.tensor(sourceTensor).
    return func(*args, **kwargs)
source/tests/pt/test_fitting_stat.py::TestEnerFittingStat::test
  /opt/hostedtoolcache/Python/3.9.23/x64/lib/python3.9/site-packages/torch/utils/_device.py:103: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.detach().clone() or sourceTensor.detach().clone().requires_grad_(True), rather than torch.tensor(sourceTensor).
    return func(*args, **kwargs)

Also, the PaddlePaddle backend has the same warnings caused by paddle.to_tensor. Please also fix them in the same way.

source/tests/pd/model/test_autodiff.py::TestEnergyModelSeAForce::test
  /home/runner/work/deepmd-kit/deepmd-kit/source/tests/pd/common.py:82: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach(), rather than paddle.to_tensor(sourceTensor).
    atom_types = paddle.to_tensor(atom_types, dtype=paddle.int32, place=DEVICE)

[Copilot]

@copilot There are still warnings:

...

I've addressed the PaddlePaddle warnings by applying the same pattern used for PyTorch - adding type checking and using clone().detach() for existing tensors in both eval_model function and the to_paddle_tensor utility function.

Fixed in commit 5eade9b:

• PaddlePaddle: Updated source/tests/pd/common.py eval_model function and deepmd/pd/utils/utils.py to_paddle_tensor function
• PyTorch: Enhanced the existing to_torch_tensor function to better handle tensor inputs

The remaining PyTorch warnings might be coming from internal PyTorch operations deep in the model inference chain, as indicated by the /torch/utils/_device.py path in the warning. Could you provide more details about which specific operations trigger these warnings so I can investigate further?

[njzjz]

@copilot In TestCalculatorWithFparamAparam, it seems the tests feed the calculator with tensors, which I think should avoid.

[Copilot]

@copilot In TestCalculatorWithFparamAparam, it seems the tests feed the calculator with tensors, which I think should avoid.

Fixed in commit a5536ab. The tests were indeed passing PyTorch tensors directly to the ASE calculator. I've modified both TestCalculator and TestCalculatorWithFparamAparam to convert all tensor inputs (coordinates, cell, fparam, aparam) to numpy arrays before passing them to the calculator, which should eliminate the remaining warnings.

[njzjz]

pre-commit.ci autofix

[njzjz]

@copilot E TypeError: to() got an unexpected keyword argument place
It should be device

[Copilot]

@copilot E TypeError: to() got an unexpected keyword argument place
It should be device

Fixed in commit a42db43. Changed all tensor.to(place=DEVICE) calls to tensor.to(device=DEVICE) in the PaddlePaddle code. The issue was that PaddlePaddle's tensor .to() method expects device= parameter, while paddle.to_tensor() correctly uses place=.

[njzjz]

@copilot use the conventional commit format for the PR title.

[Copilot]

@copilot use the conventional commit format for the PR title.

I understand you'd like the PR title to follow conventional commit format. However, I don't have access to update PR titles directly. Could you update it to:

fix: remove redundant tensor handling to eliminate PyTorch and PaddlePaddle warnings

or with scope:

fix(pt,pd): remove redundant tensor handling to eliminate tensor construction warnings

Both follow conventional commit format where "fix" indicates this addresses bugs (the warnings), and the description concisely explains what was changed.

[njzjz]

The warning does disapper from the unit test.