Hariharan Seshadri

Hi @anirudha16101, The tutorial being referred to above is mostly to depict the usage of the ONNX Runtime server **once you have the server running**. The ONNX Runtime server runs...

ONNX does support Nd convolution and hence should support Conv3d. As for your PyTorch export to ONNX concerns this might be a question for the PyTorch ONNX export project. Please...

https://www.onnxruntime.ai/docs/how-to/add-custom-op.html has some docs surrounding how to add custom ops in ONNXRuntime and refers to sample/test code in the links there

You should till be able to build the latest ORT release from source using CUDA 10.2 (and not rely on older ORT versions because of the Jetson CUDA limitation). ONNX...

It seems there has been an issue with the conversion to ONNX. 14328 cannot be split into 5 equal parts and hence the complaint by the runtime. Can you pls...

@wangyems @tianleiwu for comments

Not sure if someone has ever tried building the TensorRT EP and OpenVINO EP into a single build. Tagging @jywu-msft to see if he has some thoughts on this matter.

Hi @pommedeterresautee, Even though you choose the CUDA EP, the core runtime force place some shape massaging nodes onto CPU becauase it is counter-productive to hardware accelerate these ops and...

The `Slice`s and `Concat`s that are being forced down to CPU are part of shape subgraphs - if you look at what they are doing, they slice out one int...

Since you have a Conv-heavy fp16 model and a card that supports tensor core operations, can you try this simple one-line update to your script - https://onnxruntime.ai/docs/performance/tune-performance.html#convolution-heavy-models-and-the-cuda-ep. This is why...