Custom Ensemble approach

Hello Everyone,
I am new to pyotorch and learning quickly.
I have a dataset of images with 5 category/classes with 2 label.
For prediction of those 2 labels I plan to do an custom ensemble approach.

I want to train separate but same network(densenet-121) for each category/classes. And then the resulting model would predict out of these 2 label. I could not find a similar example for pytorch, If anyone has anyidea or could give me an concrete example would be very helpful.

For example: let’s say I have X,Y,Z,P,Q classes who have label 0, 1

Now I want train total 5 networks. densenet-121_X with the data of X, densenet-121_Y, with the data of Y and so on.
And finally somehow combine those trained model to output a single prediction of 0 or 1 given any test data from X,Y,Z,P,Q classes. I cant not figure out how to do it and not sure if it is feasible too.

I saw this example Combining Trained Models in PyTorch which is somewhat similar to my idea. But in the end instead of passing two input, I would like to pass one input to get single prediction either 0 or 1 .

Similar to bagging apporach:

instead of test sample there would images from each category
Thanks a lot in advance for help.

You could train the 5 classifiers separately as binary classifiers and try to combine them afterwards as shown in your figure.
For this you could pass all pretrained models to the linked ensemble model (maybe remove the last classification layer of each model), and add a classifier head on top of their concatenated outputs.
Would that work or are you thinking about another approach?

Thank you, you write the idea properly and that make sense to me :slight_smile: I will start working on it.
P.S. do you already have any similar example or could you give me a dummy one at least?
Getting 5 seperate trained models as .pth file is ok. But for the later steps,
removing the last classification layer, concatenate and add a classifier head on top of their concatenated outputs ? just to be sure.

Sure, we can just use the linked code as the base script:

class MyEnsemble(nn.Module):
    def __init__(self, modelA, modelB, nb_classes=10):
        super(MyEnsemble, self).__init__()
        self.modelA = modelA
        self.modelB = modelB
        # Remove last linear layer
        self.modelA.fc = nn.Identity()
        self.modelB.fc = nn.Identity()
        
        # Create new classifier
        self.classifier = nn.Linear(2048+512, nb_classes)
        
    def forward(self, x):
        x1 = self.modelA(x.clone())  # clone to make sure x is not changed by inplace methods
        x1 = x1.view(x1.size(0), -1)
        x2 = self.modelB(x)
        x2 = x2.view(x2.size(0), -1)
        x = torch.cat((x1, x2), dim=1)
        
        x = self.classifier(F.relu(x))
        return x

# Train your separate models
# ...
# We use pretrained torchvision models here
modelA = models.resnet50(pretrained=True)
modelB = models.resnet18(pretrained=True)

# Freeze these models
for param in modelA.parameters():
    param.requires_grad_(False)

for param in modelB.parameters():
    param.requires_grad_(False)

# Create ensemble model
model = MyEnsemble(modelA, modelB)
x = torch.randn(1, 3, 224, 224)
output = model(x)
4 Likes

Thanks a lot. I will try and let here know how it went :slight_smile:

Hi,
I followed the instructions and trained 7 model and adapted your code for me. But when I pass an input to the model , I get following error:

File "ModelEnsemble.py", line 162, in forward
    x = self.classifier(F.relu(x))
  File "/home/user/.local/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/user/.local/lib/python3.6/site-packages/torch/nn/modules/linear.py", line 92, in forward
    return F.linear(input, self.weight, self.bias)
  File "/home/user/.local/lib/python3.6/site-packages/torch/nn/functional.py", line 1406, in linear
    ret = torch.addmm(bias, input, weight.t())
RuntimeError: size mismatch, m1: [1 x 1053696], m2: [11648 x 7] at /pytorch/aten/src/THC/generic/THCTensorMathBlas.cu:268

All my trained models are based on densenet169, So my in_features= 1664.

Could you post the current model definition, so that I could have a look?
Based on the shape mismatch error, if looks like each feature has a shape of 150528.

Here is the code:

class MyEnsemble(nn.Module):
    def __init__(self, model_1, model_2, model_3, model_4, model_5, model_6, model_7,nb_classes):
        super(MyEnsemble,self).__init__()
        self.model_1   = model_1 
        self.model_2   = model_2 
        self.model_3  = model_3  
        self.model_4     = model_4
        self.model_5  = model_5
        self.model_6 = model_6 
        self.model_7    = model_7
        
        #Now Remove the Last layer
        self.model_1    = nn.Identity() 
        self.model_2   = nn.Identity() 
        self.model_3  = nn.Identity() 
        self.model_4     = nn.Identity()
        self.model_5  = nn.Identity() 
        self.model_6 = nn.Identity() 
        self.model_7    = nn.Identity() 
        
        
        #Now We  a head on top of those models
        # The trained model I am passing here is binary type based on densenet169
        # so I added a sequential `nn.Linear(num_ftrs, 2)` with them for output when they were trained
        # so, sould I use 2 instead of 1664 ? but used 1664 any way. 
        self.classifier = nn.Linear(1664+1664+1664+1664+1664+1664+1664,nb_classes)
    
    
    def forward(self, x):
        x1 = self.model_1(x.clone())  # clone to make sure x is not changed by inplace methods
        x1 = x1.view(x1.size(0), -1)
        
        x2 = self.model_2(x)
        x2 = x2.view(x2.size(0), -1)
        
        x3 = self.model_3(x)
        x3 = x3.view(x3.size(0), -1)
        
        x4 = self.model_4(x)
        x4 = x4.view(x4.size(0), -1)
        
        x5 = self.model_5(x)
        x5 = x5.view(x5.size(0), -1)
        
        x6 = self.model_6(x)
        x6 = x6.view(x6.size(0), -1)
        
        x7 = self.model_7(x)
        x7 = x7.view(x7.size(0), -1)
        
        
        #final
        x = torch.cat((x1, x2, x3, x4, x5, x6, x7), dim=1)
        x = self.classifier(F.relu(x))
        
        return x

It looks like you are replacing the complete model with nn.Identity layers instead of only the last linear layer.
Could you check that?

1 Like

Hi,
You are correct :frowning_face: Due to blindly following your example , I forgot that it was a dummy one.

Simple putting

self.model_1.classifier    = nn.Identity() 
......... 
...........

Solves that problems.
Thanks again :slight_smile:

2 Likes

Hi,ptrblck. I am new to pyotorch. I want to use ensemble of ConvNets to test the mnist dataset. However, I got an error as followed,
File “D:/Qw/EnsembleConvNet/mnist_Ensemble.py”, line 156, in init
self.modelA.fc = nn.Identity()

AttributeError: module ‘torch.nn’ has no attribute ‘Identity’.

type or paste code here# -*- coding: utf-8 -*-
"""
Created on Wed Jan  1 20:16:57 2020

@author: 515
"""

import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
import torch.nn.functional as F
import time

begin_time = time.clock()

# Device configuration
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
# Hyper parameters
num_epochs = 5
num_classes = 10
batch_size = 100
learning_rate = 0.001

# MNIST dataset
train_dataset = torchvision.datasets.MNIST(root='./mnist_data/',
                                           train=True, 
                                           transform=transforms.ToTensor(),
                                           download=False)

test_dataset = torchvision.datasets.MNIST(root='./mnist_data/',
                                          train=False, 
                                          transform=transforms.ToTensor())

# Data loader
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
                                           batch_size=batch_size, 
                                           shuffle=True)

test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
                                          batch_size=batch_size, 
                                         shuffle=False)

class InceptionA(nn.Module):

    def __init__(self, in_channels):
        super(InceptionA, self).__init__()
        self.branch1x1 = nn.Conv2d(in_channels, 16, kernel_size=1)

        self.branch5x5_1 = nn.Conv2d(in_channels, 16, kernel_size=1)
        self.branch5x5_2 = nn.Conv2d(16, 24, kernel_size=5, padding=2)

        self.branch3x3dbl_1 = nn.Conv2d(in_channels, 16, kernel_size=1)
        self.branch3x3dbl_2 = nn.Conv2d(16, 24, kernel_size=3, padding=1)
        self.branch3x3dbl_3 = nn.Conv2d(24, 24, kernel_size=3, padding=1)

        self.branch_pool = nn.Conv2d(in_channels, 24, kernel_size=1)

    def forward(self, x):
        branch1x1 = self.branch1x1(x)

        branch5x5 = self.branch5x5_1(x)
        branch5x5 = self.branch5x5_2(branch5x5)

        branch3x3dbl = self.branch3x3dbl_1(x)
        branch3x3dbl = self.branch3x3dbl_2(branch3x3dbl)
        branch3x3dbl = self.branch3x3dbl_3(branch3x3dbl)

        branch_pool = F.avg_pool2d(x, kernel_size=3, stride=1, padding=1)
        branch_pool = self.branch_pool(branch_pool)

        outputs = [branch1x1, branch5x5, branch3x3dbl, branch_pool]
        return torch.cat(outputs, 1)


class MyModelA(nn.Module):
    def __init__(self):
        super(MyModelA, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(88, 20, kernel_size=5)

        self.incept1 = InceptionA(in_channels=10)
        self.incept2 = InceptionA(in_channels=20)

        self.mp = nn.MaxPool2d(2)
        self.fc = nn.Linear(1408, 10)

    def forward(self, x):
        in_size = x.size(0)
        x = F.relu(self.mp(self.conv1(x)))
        x = self.incept1(x)
        x = F.relu(self.mp(self.conv2(x)))
        x = self.incept2(x)
        x = x.view(in_size, -1)  # flatten the tensor
        x = self.fc(x)
        
        return x

class MyModelB(nn.Module):
    
    def __init__(self, num_classes=10):
        super(MyModelB, self).__init__()
        self.layer1 = nn.Sequential(
                nn.Conv2d(in_channels=1, out_channels=64, kernel_size=3, stride=1, padding=1),
                nn.BatchNorm2d(64),
                nn.ReLU(),
                # stride默认为 1, padding默认为 0
                nn.Conv2d(64, 64, 3, padding=1),
                nn.BatchNorm2d(64),
                nn.ReLU(),

                nn.MaxPool2d(kernel_size=2, stride=2)
                )
        self.layer2 = nn.Sequential(
                nn.Conv2d(in_channels=64, out_channels=128, kernel_size=3, stride=1, padding=1),
                nn.BatchNorm2d(128),
                nn.ReLU(),

                nn.Conv2d(128, 128, 3, padding=1),
                nn.BatchNorm2d(128),
                nn.ReLU(),

                nn.MaxPool2d(kernel_size=2, stride=2)
                )
        self.classifer = nn.Sequential(
                nn.Linear(128*7*7,1024),
                nn.ReLU(),
                # torch.nn.Dropout(p=0.5, inplace=False)默认p=0.5
                nn.Dropout(),
                
                nn.Linear(1024, 1024),
                nn.ReLU(),
                nn.Dropout(),
                
                nn.Linear(1024, num_classes)
                )
 
    def forward(self, x):
        in_size = x.size(0) # in_size(0) 即为batch_size
        
        out = self.layer1(x)
        out = self.layer2(out)

        out = out.reshape(in_size, -1)# faltten tht out to(batch_size, 1*1*512)
        out = self.classifer(out)
        return out
        
        
class MyEnsemble(nn.Module):
    def __init__(self, modelA, modelB, nb_classes=10):
        super(MyEnsemble, self).__init__()
        self.modelA = modelA
        self.modelB = modelB
        # Remove last linear layer
        self.modelA.fc = nn.Identity()
        self.modelB.classifer = nn.Identity()
        
        # Create new classifier
        self.classifier = nn.Linear(1408+128*7*7, nb_classes)
        
    def forward(self, x):
        x1 = self.modelA(x.clone())  # clone to make sure x is not changed by inplace methods
        x1 = x1.view(x1.size(0), -1)
        x2 = self.modelB(x)
        x2 = x2.view(x2.size(0), -1)
        x = torch.cat((x1, x2), dim=1)
        
        x = self.classifier(F.relu(x))
        return x
    
modelA = MyModelA()
modelB = MyModelB()
modelA.load_state_dict(torch.load('mnist_Inception_model.ckpt'))
modelB.load_state_dict(torch.load('mnist_VGG16_BN_model.ckpt'))
model = MyEnsemble(modelA, modelB).to(device)
# Loss and optimizer
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)

# for updating learning rate
def update_lr(optimizer, lr):
    for param_group in optimizer.param_groups:
        param_group['lr'] = lr

# Train the model
total_step = len(train_loader)
curr_lr = learning_rate
for epoch in range(num_epochs):
    for i, (images, labels) in enumerate(train_loader):
        images = images.to(device)
        labels = labels.to(device)
        
        # Forward pass
        outputs = model(images)
        loss = criterion(outputs, labels)
        
        # Backward and optimize
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        
        if (i+1) % 100 == 0:
            print ('Epoch [{}/{}], Step [{}/{}], Loss: {:.4f}' 
                   .format(epoch+1, num_epochs, i+1, total_step, loss.item()))
            
    # Decay learning rate
    curr_lr = (0.95**epoch)*curr_lr
    update_lr(optimizer, curr_lr)

# Test the model
model.eval()  # eval mode (batchnorm uses moving mean/variance instead of mini-batch mean/variance)
with torch.no_grad():
    correct = 0
    total = 0
    for images, labels in test_loader:
        images = images.to(device)
        labels = labels.to(device)
        outputs = model(images)
        _, predicted = torch.max(outputs.data, 1)
        total += labels.size(0)
        correct += (predicted == labels).sum().item()
    end_time = time.clock()
    print('Test Accuracy of the model on the 10000 test images: {} %'.format(100 * correct / total))
    print('time cost: {}'.format(end_time - begin_time))

Which PyTorch version are you using? You might use an older version, where this layer hasn’t been introduced yet. Could you update to the latest stable release and rerun your script?

Hi,
As ptrblck mentioned, this is a version problem. I also got the same error. But when I updated the Pytorch to the latest version. This error was solved.

1 Like

Yes, I used the Pytorch version of 1.0, and I update to the version of 1.3, it worked.
And I have the other question::rofl:
Inception model
Test Accuracy of the model on the 10000 test images: 98.88 %
Inception Net without lr-decay, time cost: 165.26964279999999
VGG model:
Test Accuracy of the model on the 10000 test images: 99.5 %
VGG16_BN with lr-decay, time cost: 125.38929489999964
Ensemble model:
Test Accuracy of ensemble model on the 10000 test images: 99.55 %
time cost: 221.52177070000016
Is it reasonable? I have saved the pretrained model, and I trained the ensemble model, it cost more time?
Am I doing the right ensemble convs?
Waiting for your reply~:smile:

I would assume the ensemble to cost more time, as each sample has to be processed by two feature extractors now.

hi @ptrblck
i tried your solution but got this error :

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<timed exec> in <module>

/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
    539             result = self._slow_forward(*input, **kwargs)
    540         else:
--> 541             result = self.forward(*input, **kwargs)
    542         for hook in self._forward_hooks.values():
    543             hook_result = hook(self, input, result)

<ipython-input-21-1e20d3981e0b> in forward(self, x)
     14 
     15     def forward(self, x):
---> 16         x1 = self.model(x.clone())  # clone to make sure x is not changed by inplace methods
     17         x1 = x1.view(x1.size(0), -1)
     18         x2 = self.model1(x)

/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
    539             result = self._slow_forward(*input, **kwargs)
    540         else:
--> 541             result = self.forward(*input, **kwargs)
    542         for hook in self._forward_hooks.values():
    543             hook_result = hook(self, input, result)

TypeError: forward() missing 1 required positional argument: 'x2'

my model has 3 fc layers and i am using single channel images instead,so i am not using pretrained models,so i modified the ensemble code like this :

num_features = 1024
class MyEnsemble(nn.Module):
    def __init__(self, model, model1):
        super(MyEnsemble, self).__init__()
        self.model = model
        self.model1 = model1
        # vowel_diacritic
        self.fc1 = nn.Linear(num_features,11)
        # grapheme_root
        self.fc2 = nn.Linear(num_features,168)
        # consonant_diacritic
        self.fc3 = nn.Linear(num_features,7)

        
    def forward(self, x):
        x1 = self.model(x.clone())  # clone to make sure x is not changed by inplace methods
        x1 = x1.view(x1.size(0), -1)
        x2 = self.model1(x)
        x2 = x2.view(x2.size(0), -1)
        x = torch.cat((x1, x2), dim=1)
        x1 = self.fc1(x)
        x2 = self.fc2(x)
        x3 = self.fc3(x)
        return x1,x2,x3
    

where i am making mistakes?

self.model expects two inputs, while only one was passed.
What model did you pass as the first argument to MyEnsemble?

First Argument to MyEnsemble is resnet34 model modified for single channel images,here is how that model looks like :

ResNet(
  (conv1): Conv2d(1, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
  (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (relu): ReLU(inplace=True)
  (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
  (layer1): Sequential(
    (0): BasicBlock(
      (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (1): BasicBlock(
      (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (2): BasicBlock(
      (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
  )
  (layer2): Sequential(
    (0): BasicBlock(
      (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (downsample): Sequential(
        (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): BasicBlock(
      (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (2): BasicBlock(
      (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (3): BasicBlock(
      (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
  )
  (layer3): Sequential(
    (0): BasicBlock(
      (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (downsample): Sequential(
        (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (2): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (3): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (4): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (5): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
  )
  (layer4): Sequential(
    (0): BasicBlock(
      (conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (downsample): Sequential(
        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): BasicBlock(
      (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (2): BasicBlock(
      (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
  )
  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
  (fc1): Linear(in_features=512, out_features=11, bias=True)
  (fc2): Linear(in_features=512, out_features=168, bias=True)
  (fc3): Linear(in_features=512, out_features=7, bias=True)
  (avg_pooling): GeM(p=3.0000, eps=1e-06)
)

and the second argument is model1 which is also resnet34 for single channel image,and here is model1 summary :

ResNet(
  (conv1): Conv2d(1, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
  (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (relu): ReLU(inplace=True)
  (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
  (layer1): Sequential(
    (0): BasicBlock(
      (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (1): BasicBlock(
      (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (2): BasicBlock(
      (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
  )
  (layer2): Sequential(
    (0): BasicBlock(
      (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (downsample): Sequential(
        (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): BasicBlock(
      (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (2): BasicBlock(
      (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (3): BasicBlock(
      (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
  )
  (layer3): Sequential(
    (0): BasicBlock(
      (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (downsample): Sequential(
        (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (2): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (3): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (4): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (5): BasicBlock(
      (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
  )
  (layer4): Sequential(
    (0): BasicBlock(
      (conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (downsample): Sequential(
        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): BasicBlock(
      (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    (2): BasicBlock(
      (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
  )
  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
  (fc1): Linear(in_features=512, out_features=11, bias=True)
  (fc2): Linear(in_features=512, out_features=168, bias=True)
  (fc3): Linear(in_features=512, out_features=7, bias=True)
  (avg_pooling): GeM(p=3.0000, eps=1e-06)
)

can you please tell me how can i modify your ensemble code to make it work?
also i will be adding resnext50 and efficientnet models later for ensemble if this scripts works for me

Could you post the modifications to the resnets?
The code works fine using vanilla resnet34 and fixing the num_features:

num_features = 2000
class MyEnsemble(nn.Module):
    def __init__(self, model, model1):
        super(MyEnsemble, self).__init__()
        self.model = model
        self.model1 = model1
        # vowel_diacritic
        self.fc1 = nn.Linear(num_features,11)
        # grapheme_root
        self.fc2 = nn.Linear(num_features,168)
        # consonant_diacritic
        self.fc3 = nn.Linear(num_features,7)

        
    def forward(self, x):
        x1 = self.model(x.clone())  # clone to make sure x is not changed by inplace methods
        x1 = x1.view(x1.size(0), -1)
        x2 = self.model1(x)
        x2 = x2.view(x2.size(0), -1)
        x = torch.cat((x1, x2), dim=1)
        x1 = self.fc1(x)
        x2 = self.fc2(x)
        x3 = self.fc3(x)
        return x1,x2,x3
    
    
model = MyEnsemble(models.resnet34(), models.resnet34())
x = torch.randn(2, 3, 224, 224)
output = model(x)

@ptrblck you are showing me examples for 3 channel images model,but my model is designed for single channel images,here is the code i used for training single models :

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

def conv1x1(in_planes, out_planes, stride=1):
    """1x1 convolution"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)


class Bottleneck(nn.Module):
    expansion = 4
    __constants__ = ['downsample']

    def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
                 base_width=64, dilation=1, norm_layer=None):
        super(Bottleneck, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        width = int(planes * (base_width / 64.)) * groups
        # Both self.conv2 and self.downsample layers downsample the input when stride != 1
        self.conv1 = conv1x1(inplanes, width)
        self.bn1 = norm_layer(width)
        self.conv2 = conv3x3(width, width, stride, groups, dilation)
        self.bn2 = norm_layer(width)
        self.conv3 = conv1x1(width, planes * self.expansion)
        self.bn3 = norm_layer(planes * self.expansion)
        self.relu = nn.ReLU(inplace=True)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)
        out = self.relu(out)

        out = self.conv3(out)
        out = self.bn3(out)

        if self.downsample is not None:
            identity = self.downsample(x)

        out += identity
        out = self.relu(out)

        return out


class ResNet(nn.Module):

    def __init__(self, block, layers, num_classes=1000, zero_init_residual=False,
                 groups=1, width_per_group=64, replace_stride_with_dilation=None,
                 norm_layer=None):
        super(ResNet, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        self._norm_layer = norm_layer

        self.inplanes = 64
        self.dilation = 1
        if replace_stride_with_dilation is None:
            # each element in the tuple indicates if we should replace
            # the 2x2 stride with a dilated convolution instead
            replace_stride_with_dilation = [False, False, False]
        if len(replace_stride_with_dilation) != 3:
            raise ValueError("replace_stride_with_dilation should be None "
                             "or a 3-element tuple, got {}".format(replace_stride_with_dilation))
        self.groups = groups
        self.base_width = width_per_group
        self.conv1 = nn.Conv2d(1, self.inplanes, kernel_size=7, stride=2, padding=3,
                               bias=False)
        self.bn1 = norm_layer(self.inplanes)
        self.relu = nn.ReLU(inplace=True)
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        self.layer1 = self._make_layer(block, 64, layers[0])
        self.layer2 = self._make_layer(block, 128, layers[1], stride=2,
                                       dilate=replace_stride_with_dilation[0])
        self.layer3 = self._make_layer(block, 256, layers[2], stride=2,
                                       dilate=replace_stride_with_dilation[1])
        self.layer4 = self._make_layer(block, 512, layers[3], stride=2,
                                       dilate=replace_stride_with_dilation[2])
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
        # vowel_diacritic
        self.fc1 = nn.Linear(512 * block.expansion,11)
        # grapheme_root
        self.fc2 = nn.Linear(512 * block.expansion,168)
        # consonant_diacritic
        self.fc3 = nn.Linear(512 * block.expansion,7)

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)

        # Zero-initialize the last BN in each residual branch,
        # so that the residual branch starts with zeros, and each residual block behaves like an identity.
        # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677
        if zero_init_residual:
            for m in self.modules():
                if isinstance(m, Bottleneck):
                    nn.init.constant_(m.bn3.weight, 0)
                elif isinstance(m, BasicBlock):
                    nn.init.constant_(m.bn2.weight, 0)

    def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
        norm_layer = self._norm_layer
        downsample = None
        previous_dilation = self.dilation
        if dilate:
            self.dilation *= stride
            stride = 1
        if stride != 1 or self.inplanes != planes * block.expansion:
            downsample = nn.Sequential(
                conv1x1(self.inplanes, planes * block.expansion, stride),
                norm_layer(planes * block.expansion),
            )
        layers = []
        layers.append(block(self.inplanes, planes, stride, downsample, self.groups,
                            self.base_width, previous_dilation, norm_layer))
        self.inplanes = planes * block.expansion
        for _ in range(1, blocks):
            layers.append(block(self.inplanes, planes, groups=self.groups,
                                base_width=self.base_width, dilation=self.dilation,
                                norm_layer=norm_layer))

        return nn.Sequential(*layers)

    def _forward_impl(self, x):
        # See note [TorchScript super()]
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.maxpool(x)

        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)

        x = self.avgpool(x)
        x = torch.flatten(x, 1)
        x1 = self.fc1(x)
        x2 = self.fc2(x)
        x3 = self.fc3(x)
        return x1,x2,x3

    def forward(self, x):
        return self._forward_impl(x)


def _resnet(arch, block, layers, pretrained, progress, **kwargs):
    model = ResNet(block, layers, **kwargs)
    if pretrained:
        state_dict = load_state_dict_from_url(model_urls[arch],
                                              progress=progress)
        model.load_state_dict(state_dict)
    return model


class BasicBlock(nn.Module):
    expansion = 1
    __constants__ = ['downsample']

    def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
                 base_width=64, dilation=1, norm_layer=None):
        super(BasicBlock, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        if groups != 1 or base_width != 64:
            raise ValueError('BasicBlock only supports groups=1 and base_width=64')
        if dilation > 1:
            raise NotImplementedError("Dilation > 1 not supported in BasicBlock")
        # Both self.conv1 and self.downsample layers downsample the input when stride != 1
        self.conv1 = conv3x3(inplanes, planes, stride)
        self.bn1 = norm_layer(planes)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = conv3x3(planes, planes)
        self.bn2 = norm_layer(planes)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)

        if self.downsample is not None:
            identity = self.downsample(x)

        out += identity
        out = self.relu(out)

        return out

def resnet18(pretrained=False, progress=True, **kwargs):
    r"""ResNet-18 model from
    `"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    return _resnet('resnet18', BasicBlock, [2, 2, 2, 2], pretrained, progress,
                   **kwargs)


def resnet34(pretrained=False, progress=True, **kwargs):
    r"""ResNet-34 model from
    `"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    return _resnet('resnet34', BasicBlock, [3, 4, 6, 3], pretrained, progress,
                   **kwargs)


def resnet50(pretrained=False, progress=True, **kwargs):
    r"""ResNet-50 model from
    `"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    return _resnet('resnet50', Bottleneck, [3, 4, 6, 3], pretrained, progress,
                   **kwargs)


def resnet101(pretrained=False, progress=True, **kwargs):
    r"""ResNet-101 model from
    `"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    return _resnet('resnet101', Bottleneck, [3, 4, 23, 3], pretrained, progress,
                   **kwargs)


def resnet152(pretrained=False, progress=True, **kwargs):
    r"""ResNet-152 model from
    `"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    return _resnet('resnet152', Bottleneck, [3, 8, 36, 3], pretrained, progress,
                   **kwargs)


def resnext50_32x4d(pretrained=False, progress=True, **kwargs):
    r"""ResNeXt-50 32x4d model from
    `"Aggregated Residual Transformation for Deep Neural Networks" <https://arxiv.org/pdf/1611.05431.pdf>`_
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    kwargs['groups'] = 32
    kwargs['width_per_group'] = 4
    return _resnet('resnext50_32x4d', Bottleneck, [3, 4, 6, 3],
                   pretrained, progress, **kwargs)


def resnext101_32x8d(pretrained=False, progress=True, **kwargs):
    r"""ResNeXt-101 32x8d model from
    `"Aggregated Residual Transformation for Deep Neural Networks" <https://arxiv.org/pdf/1611.05431.pdf>`_
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    kwargs['groups'] = 32
    kwargs['width_per_group'] = 8
    return _resnet('resnext101_32x8d', Bottleneck, [3, 4, 23, 3],
                   pretrained, progress, **kwargs)


def wide_resnet50_2(pretrained=False, progress=True, **kwargs):
    r"""Wide ResNet-50-2 model from
    `"Wide Residual Networks" <https://arxiv.org/pdf/1605.07146.pdf>`_
    The model is the same as ResNet except for the bottleneck number of channels
    which is twice larger in every block. The number of channels in outer 1x1
    convolutions is the same, e.g. last block in ResNet-50 has 2048-512-2048
    channels, and in Wide ResNet-50-2 has 2048-1024-2048.
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    kwargs['width_per_group'] = 64 * 2
    return _resnet('wide_resnet50_2', Bottleneck, [3, 4, 6, 3],
                   pretrained, progress, **kwargs)


def wide_resnet101_2(pretrained=False, progress=True, **kwargs):
    r"""Wide ResNet-101-2 model from
    `"Wide Residual Networks" <https://arxiv.org/pdf/1605.07146.pdf>`_
    The model is the same as ResNet except for the bottleneck number of channels
    which is twice larger in every block. The number of channels in outer 1x1
    convolutions is the same, e.g. last block in ResNet-50 has 2048-512-2048
    channels, and in Wide ResNet-50-2 has 2048-1024-2048.
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    kwargs['width_per_group'] = 64 * 2
    return _resnet('wide_resnet101_2', Bottleneck, [3, 4, 23, 3],
                   pretrained, progress, **kwargs)

loading model :

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = resnet34().to(device)


from torch.nn.parameter import Parameter
def gem(x, p=3, eps=1e-6):
    return F.avg_pool2d(x.clamp(min=eps).pow(p), (x.size(-2), x.size(-1))).pow(1./p)
class GeM(nn.Module):
    def __init__(self, p=3, eps=1e-6):
        super(GeM,self).__init__()
        self.p = Parameter(torch.ones(1)*p)
        self.eps = eps
    def forward(self, x):
        return gem(x, p=self.p, eps=self.eps)       
    def __repr__(self):
        return self.__class__.__name__ + '(' + 'p=' + '{:.4f}'.format(self.p.data.tolist()[0]) + ', ' + 'eps=' + str(self.eps) + ')'
model.avg_pooling = GeM()