[MIGraphX EP] Ensure we support all inputs for MatMulInteger and ConvInteger.#21680
Conversation
… to int8 for now Allow for models with biases/full input and only check for int8 support in EP
|
Ping @tianleiwu @PeixuanZuo @ytaous Need CI run on this |
|
ping @cloudhan |
|
/azp run Windows GPU CI Pipeline, Linux CPU CI Pipeline, Linux CPU Minimal Build E2E CI Pipeline, Linux CPU x64 NoContribops CI Pipeline, Linux GPU CI Pipeline, Linux GPU TensorRT CI Pipeline, Linux OpenVINO CI Pipeline, MacOS CI Pipeline, MacOS NoContribops CI Pipeline, Windows CPU CI Pipeline |
|
/azp run Windows GPU TensorRT CI Pipeline, centos7_cpu, orttraining-linux-ci-pipeline, orttraining-linux-gpu-ci-pipeline, Android CI Pipeline, Linux DNNL CI Pipeline, iOS CI Pipeline, orttraining-win-ci-pipeline, orttraining-win-gpu-ci-pipeline, orttraining-mac-ci-pipeline |
|
Azure Pipelines successfully started running 7 pipeline(s). |
1 similar comment
|
Azure Pipelines successfully started running 7 pipeline(s). |
|
/azp run Big Models,Linux Android Emulator QNN CI Pipeline,Linux QNN CI Pipeline,ONNX Runtime Web CI Pipeline,Windows ARM64 QNN CI Pipeline,Windows GPU CUDA CI Pipeline,Windows GPU DML CI Pipeline,Windows GPU Doc Gen CI Pipeline,Windows x64 QNN CI Pipeline,onnxruntime-binary-size-checks-ci-pipeline |
|
/azp run orttraining-ortmodule-distributed |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
Azure Pipelines successfully started running 10 pipeline(s). |
|
@cloudhan Azure won't allow me to run failed jobs and it appears to be a java issue. I haven't touched anything with Java for this PR so unsure why that stage would fail |
|
@cloudhan Please let me know if there's anything else |
|
/azp run Linux MIGraphX CI Pipeline |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
Is there anything more on this? The lint failures are refencing cuda code to which I didn't touch |
|
Please run If you have not installed it before. It is like: Then fix the format according to warnings. |
|
I've run it twice with no avail. The changes referenced CUDA. I've merged main it before pushing again. I'll need to get CI kicked off again |
|
/azp run Windows GPU CI Pipeline, Linux CPU CI Pipeline, Linux CPU Minimal Build E2E CI Pipeline, Linux CPU x64 NoContribops CI Pipeline, Linux GPU CI Pipeline, Linux GPU TensorRT CI Pipeline, Linux OpenVINO CI Pipeline, MacOS CI Pipeline, MacOS NoContribops CI Pipeline, Windows CPU CI Pipeline |
|
Azure Pipelines successfully started running 7 pipeline(s). |
Still failing: After merging main, you might need run lintrunner installation again just in case some packages used in lintrunner is changed in main. If you are not sure, you can create a new conda environment and try again. |
|
Done. Had to do a reinstall. Looks like you updated ruff to 0.55 |
|
/azp run Linux MIGraphX CI Pipeline |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azp run Windows GPU CI Pipeline, Linux CPU CI Pipeline, Linux CPU Minimal Build E2E CI Pipeline, Linux CPU x64 NoContribops CI Pipeline, Linux GPU CI Pipeline, Linux GPU TensorRT CI Pipeline, Linux OpenVINO CI Pipeline, MacOS CI Pipeline, MacOS NoContribops CI Pipeline, Windows CPU CI Pipeline |
|
/azp run Big Models,Linux Android Emulator QNN CI Pipeline,Linux QNN CI Pipeline,ONNX Runtime Web CI Pipeline,Windows ARM64 QNN CI Pipeline,Windows GPU CUDA CI Pipeline,Windows GPU DML CI Pipeline,Windows GPU Doc Gen CI Pipeline,Windows x64 QNN CI Pipeline,onnxruntime-binary-size-checks-ci-pipeline |
|
Azure Pipelines successfully started running 7 pipeline(s). |
|
Azure Pipelines successfully started running 10 pipeline(s). |
|
/azp run Windows GPU TensorRT CI Pipeline,orttraining-linux-ci-pipeline,orttraining-linux-gpu-ci-pipeline,orttraining-ortmodule-distributed |
|
Azure Pipelines successfully started running 4 pipeline(s). |

… to int8 for now
Allow for models with biases/full input and only check for int8 support in EP
Description
Allows for all inputs for MatMulInteger and ConvInteger to be supported for prequantized models
Motivation and Context
Fixes issues when using prequantized models that contain weight biases