pig-pig-yang opened a new issue, #12469: URL: https://github.com/apache/tvm/issues/12469
I tried to convert Pytorch MobilenetV3 to Relay IR. But the converted Relay IR is wrong. I tried to convert to ONNX first and then convert to Relay, results are still wrong. Original structure: ![image](https://user-images.githubusercontent.com/65580829/185101766-2d7eaee7-c3e8-4ee3-8ada-f3e5121ce8ff.png) Converted Relay IR structure: ![image](https://user-images.githubusercontent.com/65580829/185101669-40a60c11-414c-47cd-9062-d913053300a1.png) As shown above, the convolution weights are connected to Mul and the branch is connected to conv2D after conversion for every Mul->Conv Node. The codes are shown below: ``` import tvm from tvm import te import tvm.relay as relay from tvm.contrib.download import download_testdata import numpy as np import os from tvm.relay.quantize.quantize import QConfig import torch import torchvision from torchvision import transforms from PIL import Image from matplotlib import pyplot as plt import netron import onnx model_name = "mobilenet_v3" model = torchvision.models.mobilenet_v3_large(pretrained=True) model = model.eval() x = torch.randn(1, 3, 224, 224) saved_path = "./mobilenet_v3.onnx" torch.onnx.export(model, x, saved_path) onnx_model = onnx.load(saved_path) input_name = "input.1" shape_list = {input_name: (1, 3, 224, 224)} mod, params = relay.frontend.from_onnx(onnx_model, shape_list) print(mod) ``` Since I am not familiar with the frontend, I don't know what causes the problem. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org