Add CUDA Graph support for the CUDA plugin EP

[tianleiwu]

Description

This PR brings CUDA graph capture/replay to the CUDA plugin execution provider so plugin-based CUDA deployments can get the same reduced CPU launch overhead that the in-tree CUDA EP already supports. It also adds the ORT framework and plugin-C-API plumbing needed to let graph-enabled plugin EPs participate safely in warmup, capture, and replay, while preserving compatibility with older plugins through version-gated fallbacks.

Summary of Changes

CUDA plugin EP runtime and allocator integration

File	Change
onnxruntime/core/providers/cuda/plugin/cuda_ep.cc	Implements plugin-side graph capture lifecycle callbacks, per-thread graph context management, graph replay, and stream selection for graph-enabled runs.
onnxruntime/core/providers/cuda/plugin/cuda_ep.h	Adds CUDA graph configuration/state to the plugin EP, including per-thread graph context ownership.
onnxruntime/core/providers/cuda/plugin/cuda_graph_plugin.cc	Adds CudaGraphSet/CudaGraphManager to own captured graphs and coordinate warmup, capture, and replay by annotation ID.
onnxruntime/core/providers/cuda/plugin/cuda_graph_plugin.h	Declares the new graph manager types and graph-related constants.
onnxruntime/core/providers/cuda/plugin/cuda_stream_plugin.cc	Adds external-stream wrapping so graph-enabled runs can reuse the thread’s graph stream without taking ownership of it.
onnxruntime/core/providers/cuda/plugin/cuda_stream_plugin.h	Declares the external-stream initialization path and stream ownership tracking.
onnxruntime/core/providers/cuda/plugin/cuda_ep_factory.cc	Parses enable_cuda_graph and min_num_runs_before_cuda_graph_capture provider/session options for the plugin EP.
onnxruntime/core/providers/cuda/plugin/cuda_mempool_allocator_plugin.cc	Updates allocator behavior needed for CUDA native mempool compatibility during graph capture/replay.
onnxruntime/core/providers/cuda/plugin/cuda_kernel_adapter.h	Adjusts plugin kernel/device helpers used by the graph-enabled execution path.
onnxruntime/core/providers/cuda/plugin/cuda_plugin_utils.h	Adds supporting helpers used by the plugin CUDA graph flow.

ORT framework and plugin API support for graph replay

File	Change
include/onnxruntime/core/session/onnxruntime_ep_c_api.h	Documents and extends the plugin EP contract for graph-enabled runs, including replay behavior relative to OnRunStart/OnRunEnd.
include/onnxruntime/core/framework/execution_provider.h	Adds graph-capture node-assignment policy support to the execution provider interface.
onnxruntime/core/session/inference_session.cc	Generalizes the session replay path and warmup/capture retry loop so ORT can drive graph replay for graph-capable EPs.
onnxruntime/core/session/inference_session.h	Updates replay-related messaging and supporting declarations for the new run flow.
onnxruntime/core/framework/session_state.cc	Makes device-stream collection reuse thread-affine so warmup/capture/replay reuse stays on the owning thread.
onnxruntime/core/framework/session_state.h	Adds supporting state for the thread-affine stream collection pool.
onnxruntime/core/session/plugin_ep/ep_plugin_provider_interfaces.cc	Bridges the new graph callbacks, hardens validation of plugin graph support, and exposes effective plugin provider options gathered from session config.
onnxruntime/core/session/plugin_ep/ep_plugin_provider_interfaces.h	Stores provider options and declares the new accessor/graph bridge behavior.
onnxruntime/core/providers/webgpu/webgpu_execution_provider.h	Aligns graph-capture policy support with the new execution-provider interface.
onnxruntime/core/providers/js/js_execution_provider.h	Aligns graph-capture policy support with the new execution-provider interface.

Tests and validation coverage

File	Change
onnxruntime/test/python/transformers/test_cuda_plugin_ep.py	Adds end-to-end CUDA graph tests for warmup/capture/replay, replay after input updates, CUDA mempool mode, multiple graph annotation IDs, multi-GPU/device-id coverage, and a simple Add model.

Documentation

File	Change
docs/cuda_plugin_ep/cuda_graph_for_cuda_plugin.md	Adds a dedicated design/implementation document covering architecture, lifecycle, allocator interaction, concurrency, and verification guidance.
docs/cuda_plugin_ep/cuda_plugin_ep_design.md	Updates the broader plugin EP design doc to reflect that CUDA graph support is implemented and documents the framework-level changes.
docs/cuda_plugin_ep/QUICK_START.md	Updates quick-start/testing guidance and removes the outdated “no CUDA Graph support” limitation.

Testing

• Build ONNX Runtime with onnxruntime_BUILD_CUDA_EP_AS_PLUGIN=ON, install the generated wheel, and deploy the CUDA plugin shared library as described in docs/cuda_plugin_ep/QUICK_START.md.
• Run python onnxruntime/test/python/transformers/test_cuda_plugin_ep.py.
• Pay particular attention to the new CUDA graph scenarios in that suite: warmup/capture/replay, replay after in-place input updates, CUDA mempool mode, multiple gpu_graph_id captures, and the second-device path when multiple GPUs are available.
• Verify backward compatibility by confirming older plugins still load safely through the version-gated graph callback bridge, and that graph-disabled runs continue through the normal execution path.

Motivation and Context

The CUDA plugin EP exists to decouple CUDA EP delivery from core ONNX Runtime releases, but that model only works well if important runtime optimizations are also available through the plugin path. CUDA graph replay is one of the highest-value CUDA execution optimizations because it eliminates repeated kernel-launch overhead after capture, especially for steady-state inference workloads.

Supporting that in the plugin EP required more than adding plugin-local capture code. ORT also needed a framework-level replay flow that works for plugin EPs, a plugin C API contract for graph support and node-assignment policy, and thread-affine stream reuse so captured graph resources and stream wrappers are not reused across unrelated threads. This PR packages those pieces together and documents the resulting behavior for future plugin EP work. It also depends on earlier plugin allocator work so warmup can stabilize allocations before capture begins.

Checklist

• Tests added/updated
• Documentation updated (if applicable)
• No breaking changes (or documented in description)

[Copilot]

Pull request overview

This PR adds CUDA Graph capture/replay support to the CUDA plugin Execution Provider (EP) by extending the OrtEp C API and teaching ORT core to orchestrate warm-up, capture, and replay in a provider-agnostic way.

Changes:

• Extend the OrtEp C API and IExecutionProvider interface to support graph capture/replay capability querying, replay, and assignment-policy-driven validation.
• Update InferenceSession to validate and manage graph capture generically (policy-driven validation, bounded warm-up recursion, replay fast-path).
• Implement CUDA Graph capture/replay in the CUDA plugin EP (per-thread graph stream/context) and add Python test coverage + docs.

Reviewed changes

Copilot reviewed 27 out of 27 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
onnxruntime/test/python/transformers/test_cuda_plugin_ep.py	Adds CUDA Graph capture/replay tests for the CUDA plugin EP.
onnxruntime/core/session/plugin_ep/ep_plugin_provider_interfaces.h	Adds graph capture/replay and provider-options overrides to PluginExecutionProvider.
onnxruntime/core/session/plugin_ep/ep_plugin_provider_interfaces.cc	Extracts provider options from session config; bridges graph capture/replay calls to plugin via OrtEp v1.26 callbacks.
onnxruntime/core/session/inference_session.h	Adds bounded warm-up recursion constant and a RunImpl helper with depth tracking.
onnxruntime/core/session/inference_session.cc	Makes graph capture validation provider-agnostic via policy; adds bounded recursive warm-up/capture behavior and replay fast-path.
onnxruntime/core/providers/webgpu/webgpu_execution_provider.h	Declares graph-capture assignment policy for WebGPU EP.
onnxruntime/core/providers/webgpu/ep/ep.h	Adds OrtEp graph capture/replay callback declarations for WebGPU wrapper EP.
onnxruntime/core/providers/webgpu/ep/ep.cc	Wires OrtEp graph capture/replay callbacks through to the underlying IExecutionProvider.
onnxruntime/core/providers/js/js_execution_provider.h	Declares graph-capture assignment policy for JS EP.
onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h	Declares graph-capture assignment policy for DML EP.
onnxruntime/core/providers/cuda/plugin/cuda_stream_plugin.h	Adds external-stream handle init support and stream ownership tracking.
onnxruntime/core/providers/cuda/plugin/cuda_stream_plugin.cc	Implements external-stream init; avoids destroying external streams; avoids synchronizing while stream is actively capturing.
onnxruntime/core/providers/cuda/plugin/cuda_plugin_utils.h	Adds a TLS “is capturing” flag to avoid illegal CUDA calls during capture.
onnxruntime/core/providers/cuda/plugin/cuda_kernel_adapter.h	Skips cudaSetDevice() during capture to avoid CUDA capture restrictions.
onnxruntime/core/providers/cuda/plugin/cuda_graph_plugin.h	New plugin-side CUDA graph manager API (capture/end/replay, per-annotation storage).
onnxruntime/core/providers/cuda/plugin/cuda_graph_plugin.cc	Implements plugin-side CUDA graph lifecycle and replay.
onnxruntime/core/providers/cuda/plugin/cuda_ep.h	Adds CUDA graph config knobs and per-thread graph context plumbing for the plugin EP.
onnxruntime/core/providers/cuda/plugin/cuda_ep.cc	Implements plugin EP graph capture/replay callbacks and per-thread graph stream/context management.
onnxruntime/core/providers/cuda/plugin/cuda_ep_factory.cc	Parses enable_cuda_graph and warm-up run count from session config/provider options.
onnxruntime/core/providers/cuda/cuda_execution_provider.h	Adds graph-capture assignment policy override for in-tree CUDA EP.
include/onnxruntime/core/session/onnxruntime_ep_c_api.h	Adds OrtEp v1.26 graph capture/replay callbacks + node-assignment-policy enum and docs.
include/onnxruntime/core/session/onnxruntime_cxx_inline.h	Adds Env::CopyTensor() convenience wrapper over CopyTensors().
include/onnxruntime/core/session/onnxruntime_cxx_api.h	Declares Env::CopyTensor() in the public C++ API.
include/onnxruntime/core/framework/execution_provider.h	Adds GetGraphCaptureNodeAssignmentPolicy() virtual with default strict policy.
docs/cuda_plugin_ep/QUICK_START.md	Updates plugin quick start/prereqs; removes outdated “no CUDA Graph” limitation.
docs/cuda_plugin_ep/cuda_plugin_ep_design.md	Updates design doc to reflect implemented CUDA Graph support and framework/API changes.
docs/cuda_plugin_ep/cuda_graph_for_cuda_plugin.md	New detailed design/implementation doc for CUDA Graph in the plugin EP.

---

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

[Copilot]

Pull request overview

Copilot reviewed 28 out of 28 changed files in this pull request and generated 4 comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 30 out of 30 changed files in this pull request and generated 3 comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 24 out of 24 changed files in this pull request and generated 1 comment.

---

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