TensorRT_Seg_LW_Refinenet icon indicating copy to clipboard operation
TensorRT_Seg_LW_Refinenet copied to clipboard

failed to parse ONNX model from unet.onnx

Open chegnyanjun opened this issue 5 years ago • 3 comments

It is a very great work, and i want to use it in my work, but when i load my unet.onnx model, It sends out the problem:Could not open file unet.onnx.Failed to parse ONNX model from unet.onnx.And i don't know why it would lead to such problem.Beside, I am use tensorrt6.0.1.5,Does it maybe the problem of tensorrt edition?

chegnyanjun avatar Nov 15 '19 06:11 chegnyanjun

hello, did you already solve the problem about nn.usample layer export onnx but tensorRT cant't parse it??

rainbow520lxr avatar Nov 22 '19 05:11 rainbow520lxr

@rainbow520lxr,Thanks to get your reply,I have not solve the probelm because the tensorrt doesn't supports the nn.upsample() layer,Do you have any good ideas?Tje tensorrt6.0.1.5 seems not support write plugin layer in python when parse the onnx model. when parse the caffe model ,you can write your plugin which the layers tensorrt does not support.It seems the tensorrt6.0.1.5 not supports very well about the pytorch.Hope for your reply,tkx.

chegnyanjun avatar Nov 22 '19 06:11 chegnyanjun

@rainbow520lxr,Thanks to get your reply,I have not solve the probelm because the tensorrt doesn't supports the nn.upsample() layer,Do you have any good ideas?Tje tensorrt6.0.1.5 seems not support write plugin layer in python when parse the onnx model. when parse the caffe model ,you can write your plugin which the layers tensorrt does not support.It seems the tensorrt6.0.1.5 not supports very well about the pytorch.Hope for your reply,tkx.

use pytorch0.4.1+tensorrt5.0

ASONG0506 avatar Jan 09 '20 10:01 ASONG0506