Adding an argument to ResNet basic block

stef427 · December 30, 2020, 8:58pm

I am modifying ResNet to train some variables emb, which depends on some input meta data. I have issues passing the meta data down to my user defined layer in ResNet.

emb is Category Embedding in the diagram, meta is “Dress” in the diagram

Main function:
outputs = model(inputs)>> outputs = model(inputs, meta), model here is ResNet 18
ResNet forward function,
def forward(self, x, meta):
x = self.conv1(x)
x = self.bn1(x)
x = self.relu(x)
x = self.layer1(x, meta)
BasicBlock forward function:
def forward(self, x, meta):

I am getting errors that don’t make sense to me. The above code gives me error message
x = self.layer1(x, meta)
File “/home/user/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 493, in call
result = self.forward(*input, **kwargs)
TypeError: forward() takes 2 positional arguments but 3 were given

But if I remove the last argument meta and run this instead, x = self.layer1(x), then I get this error

Traceback (most recent call last):
File “main.py”, line 202, in
main()
File “main.py”, line 153, in main
outputs = model(inputs, meta)
File “/home/user/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 493, in call
result = self.forward(*input, **kwargs)
File “/…/models/resnet.py”, line 471, in forward
x = self.layer1(x)
File “/home/user/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 493, in call
result = self.forward(*input, **kwargs)
File “/home/user/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/container.py”, line 92, in forward
input = module(input)
File “/home/user/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 493, in call
result = self.forward(*input, **kwargs)
TypeError: forward() missing 1 required positional argument: ‘meta’

This is the code my mod is based on,

github.com

Jongchan/attention-module/blob/master/MODELS/model_resnet.py

import torch
import torch.nn as nn
import torch.nn.functional as F
import math
from torch.nn import init
from .cbam import *
from .bam import *

def conv3x3(in_planes, out_planes, stride=1):
    "3x3 convolution with padding"
    return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
                     padding=1, bias=False)

class BasicBlock(nn.Module):
    expansion = 1

    def __init__(self, inplanes, planes, stride=1, downsample=None, use_cbam=False):
        super(BasicBlock, self).__init__()
        self.conv1 = conv3x3(inplanes, planes, stride)
        self.bn1 = nn.BatchNorm2d(planes)

This file has been truncated. show original

ptrblck · January 9, 2021, 7:57am

Could you post a code snippet to reproduce this issue, as the linked code doesn’t contain the modification including the meta information?