Skip to content

[Performance] EPCTX file check doesn't happen for default model path #26294

@preetha-intel

Description

@preetha-intel

Describe the issue

The check if an Epctx file alreay exists, is carried out at the beginning of session creation even before graph_partitioning is called to fail early when session_options.epctx_file_path key is set.

As per this PR :
8705c68#r167792340

The check is not happening for default model paths. If epctx file path is not explicitly set then EP compile happens and failure for a exisiting epctx file is thrown in dump epctx model.

@adrianlizarraga I have added the comments part of the PR as well

To reproduce

onnxruntime_perf_test.exe -v -m times -r 1 -C "ep.context_enable|1 ep.context_embed_mode|0" -i "device_type|CPU" <path_to_model>

During first run epctx.onnx model will be generated.

When the same command is run again, the process should fail early that epctx.onnx model exists.

Urgency

No response

Platform

Windows

OS Version

Windows

ONNX Runtime Installation

Built from Source

ONNX Runtime Version or Commit ID

8705c68

ONNX Runtime API

C++

Architecture

X64

Execution Provider

Default CPU

Execution Provider Library Version

No response

Model File

No response

Is this a quantized model?

No

Metadata

Metadata

Assignees

No one assigned

    Labels

    performanceissues related to performance regressions

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions