Question about the GoogLeNet relay

Questions
#1

I want to test the function “CombineParallelConv2D”, so I use the GoogLenet. The below is my code. import numpy as np

import tvm

from tvm import relay

from tvm.autotvm.graph_tuner import DPTuner

from tvm.contrib import graph_runtime

import torch

import torchvision

from torchvision import models

googlenet = models.googlenet(pretrained=True)

model = googlenet.eval()

model.dropout = torch.nn.Dropout(p = 0)

input_shape = [1, 3, 224, 299]

input_data = torch.randn(input_shape)

scripted_model = torch.jit.trace(model, input_data).eval()

input_name = ‘img’

shape_list = [(input_name, input_shape)]

mod, params = relay.frontend.from_pytorch(scripted_model, shape_list)

seq = tvm.transform.Sequential(

[

    relay.transform.FoldConstant(),

    relay.transform.EliminateCommonSubexpr(),

    relay.transform.FuseOps(),

    relay.transform.CombineParallelConv2D(3),

]

) mod1 = seq(mod)

And my error is :

“Check failed: (idx < data_.size() && data_[idx].second != 0) is false: Attribute TOpPattern has not been registered for nn.batch_norm”