[VitisAI] bugfix model_clone optimization

[wcy123]

Description

It is related to #25320 #23979. Enable tensor raw data sharing for externalized tensor proto with kTensorProtoMemoryAddressTag

Motivation and Context

With #25320 #23979, all initialized tensor protos are associated with OrtValue, VitisiAI EP need to adapt to this change.

[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).

[adrianlizarraga]

Hi @wcy123,
Thank you for tracking this down. Could you please try evaluating this branch to see if it resolves the issue: #25626

We would like to avoid exposing the kTensorProtoMemoryAddressTag functionality to EPs as that should be internal to onnxruntime. Thank you.

[jywu-msft]

Hi @wcy123, Thank you for tracking this down. Could you please try evaluating this branch to see if it resolves the issue: #25626

We would like to avoid exposing the kTensorProtoMemoryAddressTag functionality to EPs as that should be internal to onnxruntime. Thank you.

yes. let's not merge this PR. pls validate the general fix works for you. thank you!

[wcy123]

Hi @wcy123, Thank you for tracking this down. Could you please try evaluating this branch to see if it resolves the issue: #25626
We would like to avoid exposing the kTensorProtoMemoryAddressTag functionality to EPs as that should be internal to onnxruntime. Thank you.

yes. let's not merge this PR. pls validate the general fix works for you. thank you!

Based on our comprehensive local testing, we observed that while PR #25626 successfully addresses issues for some models, several others continue to experience the same problems. Therefore, this current PR remains necessary to fully resolve the outstanding issues.

cc Guggilla, Nithin Kumar Bader Adam.

[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]

There's another fix in progress: #25652

[yuslepukhin]

if (tensor_proto->data_location() == ONNX_NAMESPACE::TensorProto_DataLocation::TensorProto_DataLocation_EXTERNAL) {

utils::HasExternalData() or utils::HasExternalDataInMemory()

This is available from core/framework/tensorprotoutil.h

Creating and reading external data is encoded in ExternalData class.

The same utils have Unpack() methods that handle all kinds of data within TensorProto

---

Refers to: onnxruntime/core/providers/vitisai/imp/tensor_proto.cc:18 in 1109d03. [](commit_id = 1109d03, deletion_comment = False)

[yuslepukhin]

/azp run Test Linux CUDA x64 Release,Test Linux TensorRT x64 Release,Windows GPU Doc Gen CI Pipeline, windows_x64_debug/build_x64_debug (pull_request),web_Debug/build_onnxruntime_web, web_Debug / build_onnxruntime_web

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[yuslepukhin]