Encryption does not work with trt_dump_ep_context_model

[BengtGustafsson]

Describe the issue

If trt_dump_ep_context_model is set to optimize setup performance the decryption step is not done, the code in Status TensorRTCacheModelHandler::GetEpContextFromGraph(const GraphViewer& graph_viewer) only looks for the unencrypted file, while if encryption is enabled and an encryption library is provided its encrypt function is called to encrypt the function (and the .encrypted extension added).

To reproduce

Set the trt_dump_ep_context_model flag and set up encryption, you get an encrypted file but when read back the code only looks for the unencrypted file.

Urgency

We have already given up on being able to use your encryption for our next release, so it is not that urgent for us.

Platform

Windows

OS Version

10 & 11

ONNX Runtime Installation

Built from Source

ONNX Runtime Version or Commit ID

1.20

ONNX Runtime API

C++

Architecture

X64

Execution Provider

TensorRT

Execution Provider Library Version

CUDA 11.8, TrT 10.4.0.26

[jingyanwangms]

Hi @BengtGustafsson thanks for reporting this. Can you please share a reproducible example? Especially what's the engine_decryption_lib_path_ do you use

[chilo-ms]

@BengtGustafsson TRT EP doesn't support EP Context feature for the decryption workflow due to there is no requests for it.
Just wondering have you been using the decryption feature in the releases for a while? If so, we can discuss internally and help add the EP Context feature for it if needed.

[github-actions]

This issue has been automatically marked as stale due to inactivity and will be closed in 30 days if no further activity occurs. If further support is needed, please provide an update and/or more details.