Enable get dropout_mask

[chunhuanMeng]

Introduces a new utility for getting dropout masks on XPU devices, primarily for testing purposes.

Dropout Mask Utility

• Added the _fill_mem_eff_dropout_mask_ function in Attention.cpp, which generates a dropout mask using XPU kernels and copies it into the input tensor. This utility is intended for testing and does not prioritize performance.

[Copilot]

Pull Request Overview

This PR introduces a utility function for generating dropout masks on XPU devices, primarily for testing purposes. The implementation adds a new native function that uses XPU kernels to create dropout masks.

Key Changes:

• Added _fill_mem_eff_dropout_mask_ native function declaration in YAML configuration
• Implemented the dropout mask generation utility in Attention.cpp using XPU dropout kernels

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
yaml/native/native_functions.yaml	Declares the new _fill_mem_eff_dropout_mask_ function with XPU dispatch and nondeterministic_seeded tag
src/ATen/native/transformers/Attention.cpp	Implements the dropout mask utility and adds necessary includes for XPU dropout kernels

---

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

[github-actions]

Performance outliers, please check!

• 🔴 [-1, 80%), should be regression

Category	Model	Target vs. Baseline [Eager]	Target vs. Baseline [Inductor]
timm_models_bfloat16_training	beit_base_patch16_224	0.999691	0.781541