Is there any mehtod for converting mobilenet0.25 to TensorRT engine?

[G-Bong]

Hello all.

I am currently working with MobiletNet0.25 to convert the model to TensorRT.

I tried conversion with TensorRT environment supported by Nvidia.

However I couldn't get success.

Do you guys have any method?

Thanks in advance.

[gan3sh500]

@G-Bong I have a fork with TRT engine conversion added. https://github.com/gan3sh500/retinaface-pytorch

[G-Bong]

@gan3sh500 Thank you so much. Would you mind if I study your code for the TensorRT skill ? :)

[gan3sh500]

@G-Bong go ahead. It is mostly modified from the links I mention in the code. That code is just specific for single input, single output.

[G-Bong]

@gan3sh500 Thanks a lot

[G-Bong]

@gan3sh500 Sir, I have a question. When I tried to convert the mobilenet to tensorrt, the interpolate layer occurs the error for onnx to tensorrt. Didn't you meet the error? If you solved, how did it??

[yumaofan]

@G-Bong I have just solved it, The question occurs when PyTorch F.interpolate module converts to ONNX-Resize layer, The interpolate size param can't mapping with ONNX-Resize scale_factor param. So, you can attempt to set scale_factor=2 in interpolate directly. like this: up3 = F.interpolate(output3, scale_factor=2, mode="nearest") up2 = F.interpolate(output2, scale_factor=2, mode="nearest") BTW, The net input size must be the same

[HoangTienDuc]

hello @gan3sh500 can you share your onnx2tensorrt conversion?

[QuantumLiu]

hello @gan3sh500 can you share your onnx2tensorrt conversion?

Just using TensorRt>7.1

[HoangTienDuc]

@QuantumLiu why and where is tensorrt Retinaface in tensorrt >7.1?

[QuantumLiu]

@QuantumLiu why and where is tensorrt Retinaface in tensorrt >7.1?

I mean use trtexec of TensorRt>7.1 to generate engine file from onnx file. TensorRT 7.1 surpport F.interpolate function.