[CUDA] Deprecate SkipLayerNorm strict mode

[tianleiwu]

Description

Removes the deprecated enable_skip_layer_norm_strict_mode CUDA provider option and its associated "strict mode" code path from the CUDA SkipLayerNorm kernel.

Strict mode previously routed SkipLayerNorm through the LayerNormalization kernel to gain fp32-accumulation accuracy at the cost of performance. After #28682 made SkipLayerNorm/EmbedLayerNorm CUDA kernels always accumulate in fp32, the strict-mode path is redundant: the default kernel already provides the same accuracy with better performance. This change deletes the now-dead branch and the plumbing that fed it.

Key Changes

• skip_layer_norm.cc: Remove the strict_ branch that called HostApplyLayerNorm; always launch LaunchSkipLayerNormKernel. Drop the now-unused layer_norm_impl.h include and the input/skip same-shape strict-mode shape check.
• skip_layer_norm.h: Remove the strict_ member.
• cuda_execution_provider.h: Remove IsSkipLayerNormInStrictMode().
• cuda_kernel_adapter.h: Remove GetCudaKernelAdapterSkipLayerNormStrictMode() shim.
• cuda_provider_options.h: Keep enable_skip_layer_norm_strict_mode field for ABI/back-compat but mark it deprecated and ignored.
• skiplayernorm_op_test.cc: Drop the redundant strict-mode test passes; tests now run a single default path.

The provider option is retained (ignored) to preserve backward compatibility — existing configs that set it continue to work without error.

Motivation

Follow-up cleanup to #28682, which switched the CUDA kernels to fp32 accumulation, making strict mode obsolete.

Testing

• skiplayernorm_op_test.cc covers fp16/fp32/bf16 default path; strict-mode passes removed.

[Copilot]

Pull request overview

This PR removes the deprecated CUDA SkipLayerNorm “strict mode” execution path and the associated provider-option plumbing, since SkipLayerNorm now always uses fp32 accumulation (making strict mode redundant). It also updates tests to stop running separate strict-mode passes and marks the option as deprecated/ignored for backward compatibility.

Changes:

• Remove strict-mode branching in the CUDA SkipLayerNorm kernel and delete the strict_ member/config accessors.
• Remove the CUDA plugin adapter shim used to read strict-mode config.
• Update CUDA provider options documentation to indicate the strict-mode option is deprecated, and simplify tests to a single path.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
onnxruntime/contrib_ops/cuda/bert/skip_layer_norm.cc	Removes strict-mode routing; always launches SkipLayerNorm kernel.
onnxruntime/contrib_ops/cuda/bert/skip_layer_norm.h	Drops the strict_ member from the kernel class.
onnxruntime/core/providers/cuda/cuda_execution_provider.h	Removes IsSkipLayerNormInStrictMode() accessor.
onnxruntime/core/providers/cuda/plugin/cuda_kernel_adapter.h	Removes the plugin shim accessor for strict-mode.
include/onnxruntime/core/providers/cuda/cuda_provider_options.h	Marks enable_skip_layer_norm_strict_mode as deprecated/ignored for back-compat.
onnxruntime/test/contrib_ops/skiplayernorm_op_test.cc	Removes strict-mode test runs; keeps only default path execution.

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated no new comments.

[kunal-vaishnavi]

We should also remove it from here.