fix shape inference error for ep context nodes

[wcy123]

Description

To support writing an Execution Provider (EP) using the new EP ABI introduced in #24887, this PR adds value info for EP Context nodes to prevent shape inference errors during Graph::Resolve.

Motivation and Context

When creating a new EP Context node whose input is the output of another EP Context node, Graph::Resolve fails to set the type for the new node's arguments. This is because EP Context nodes do not have a TypeAndShapeInferenceFunction defined, as shown here:

onnxruntime/onnxruntime/core/graph/contrib_ops/contrib_defs.cc

Lines 3289 to 3337 in 5fdd4e4

	ONNX_CONTRIB_OPERATOR_SCHEMA(EPContext)
	.SetDomain(kMSDomain)
	.SinceVersion(1)
	.SetDoc("Onnx node container for EP context.")
	.Attr(
	"main_context",
	"Usually each single EPContext associate with a graph partition."
	"But for some case like QNN, it has single EPContext contains all partitions."
	"In that case, the node with ep_cache_context should set main_context=1. Other nodes set main_context=0 and skip ep_cache_context."
	"The path is relative to this Onnx file. Default is 1.",
	AttributeProto::INT,
	static_cast<int64_t>(1))
	.Attr(
	"ep_cache_context",
	"payload of the execution provider context if embed_mode=1, or path to the context file if embed_mode=0.",
	AttributeProto::STRING,
	OPTIONAL_VALUE)
	.Attr(
	"embed_mode",
	"1: indicate ep_cache_context is the context content. 0: indicate ep_cache_context is the file path to the context content."
	"The path is relative to this Onnx file. Default is 1.",
	AttributeProto::INT,
	static_cast<int64_t>(1))
	.Attr(
	"ep_sdk_version",
	"(Optional) SDK version used to convert the model.",
	AttributeProto::STRING,
	OPTIONAL_VALUE)
	.Attr(
	"onnx_model_filename",
	"(Optional) Filename of the original ONNX model.",
	AttributeProto::STRING,
	OPTIONAL_VALUE)
	.Attr(
	"hardware_architecture",
	"(Optional) Hardware architecture.",
	AttributeProto::STRING,
	OPTIONAL_VALUE)
	.Attr(
	"partition_name",
	"(Optional) partitioned graph name.",
	AttributeProto::STRING,
	OPTIONAL_VALUE)
	.Attr(
	"source",
	"(Optional) the source used to generate the engine/context cache file. Ort EP or native SDK tool chain",
	AttributeProto::STRING,
	OPTIONAL_VALUE)
	.Attr("notes", "(Optional) Some notes for the model", AttributeProto::STRING, OPTIONAL_VALUE)

As a result, an exception is thrown during shape inference:

onnxruntime/onnxruntime/core/graph/graph.cc

Line 2964 in 5fdd4e4

Status status(ONNXRUNTIME, onnxruntime::common::StatusCode::FAIL,

Specifically:

EP Context nodes lack TypeAndShapeInferenceFunction, so onnx_inferred_type is unavailable.
existing_type is nullptr due to the logic in:

onnxruntime/onnxruntime/core/session/ep_plugin_provider_interfaces.cc

Line 279 in 9de58ac

auto node_arg = std::make_unique<NodeArg>(input_name, /*p_arg_type*/ nullptr); // Graph.Resolve() sets type.

onnxruntime/onnxruntime/core/session/ep_plugin_provider_interfaces.cc

Line 285 in 9de58ac

auto node_arg = std::make_unique<NodeArg>(output_name, /*p_arg_type*/ nullptr); // Graph.Resolve() sets type.

Implementation

This PR attempts to add type information to EP Context nodes with best effort, ensuring that Graph::Resolve can proceed without errors even when type inference is not explicitly defined.

[HectorSVC]

/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline,Windows x64 QNN CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 5 pipeline(s).

[Copilot]

Pull Request Overview

This PR fixes shape inference errors for EP Context nodes by adding type information during their creation. The fix ensures that when EP Context nodes are chained (where one node's output becomes another's input), the Graph::Resolve process can proceed without throwing exceptions due to missing type information.

• Adds fused_nodes parameter to ConvertEpContextNodes function to access original type information
• Retrieves type information from the original fused graph's NodeArgs instead of setting them to nullptr
• Updates function signature and call site to pass the additional parameter

[HectorSVC]

/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline,Windows x64 QNN CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 5 pipeline(s).

[HectorSVC]

[snnn]

Hi there! We haven't cut the release branch for this version yet, so I'm removing the release:1.23.0 label for now to keep things tidy. Thanks so much for your contribution! We'll make sure this gets included when the release is prepared. 🤖