microsoft / semantic-kernel

Integrate cutting-edge LLM technology quickly and easily into your apps
https://aka.ms/semantic-kernel
MIT License
22.12k stars 3.3k forks source link

Python: Fix Onnx Connector Memory Problem with Onnx #9716

Open nmoeller opened 1 week ago

nmoeller commented 1 week ago

Onnx currently faces memory issues when dividing functionality among multiple methods

Motivation and Context

I was experiencing weird non reproducible memory issues with connector when using phi-3 vision, after tracing the memory it turned out there are some issues when Parameters & Generation are not in the same function. I am already in contact with the PG to adress the issue also in onnx.

There seems to be a memory problem with pybind, because the parameters show a non deterministc behavior, but they should determistic.

To fix the current problem i've decided to merge the Parameter Method and the Generation Method.

Description

Contribution Checklist

markwallace-microsoft commented 1 week ago

Python Unit Test Overview

Tests Skipped Failures Errors Time
2869 4 :zzz: 0 :x: 0 :fire: 1m 20s :stopwatch:
nmoeller commented 6 days ago

@TaoChenOSU Anything I can do here ? I only see that there is a timeout, do you experience this issue in other pipelines aswell ? Or is it just me ?

TaoChenOSU commented 2 days ago

@TaoChenOSU Anything I can do here ? I only see that there is a timeout, do you experience this issue in other pipelines aswell ? Or is it just me ?

Hi @nmoeller, our integration test pipeline is experiencing some issues. We are working on fixing it. Will merge this PR once it's fixed.