[API-Compat] paddle.Tensor.requires_grad is added and tested

[LLSGYN]

PR Category

User Experience

PR Types

New features

Description

This PR adds a requires_grad property to Paddle Tensors to provide compatibility and improve user experience.

What's added:

• Added requires_grad property to Tensor class across all execution modes (dynamic graph, static graph, and PIR)
• The property provides a convenient way to check and set gradient computation requirements
• requires_grad is the inverse of stop_gradient (requires_grad=True means stop_gradient=False)
• Added comprehensive unit tests to verify the functionality

Key features:

• Getter: Returns not self.stop_gradient, allowing users to check if gradient computation is enabled
• Setter: Accepts boolean values and sets self.stop_gradient = not value
• Type checking: Raises TypeError if non-boolean values are provided
• Documentation: Includes detailed docstrings with usage examples

Files modified:

• python/paddle/base/dygraph/math_op_patch.py - Dynamic graph mode support
• python/paddle/base/layers/math_op_patch.py - Static graph mode support
• python/paddle/pir/math_op_patch.py - PIR mode support
• test/legacy_test/test_tensor_requires_grad.py - Comprehensive test suite

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[zhwesky2010]

LGTM