model.py

import torch
import torchvision

from torch import nn
from torchvision.models import efficientnet_b0, EfficientNet_B0_Weights
from torchvision.models._api import WeightsEnum
from torch.hub import load_state_dict_from_url

def create_effnetb0_model(num_classes:int=10,
                          seed:int=42):
  """Creates an EficientNetB0 feature extractor model and transforms.

  Args:
    num_classes (int, optional): number of classes in the classifier head. Defaults to 10.
    seed (int, optional): random seed value. Defaults to 42.

  Returns:
    model (torch.nn.Module): EffNetB0 feature extractor model.
    transforms (torchvision.transforms): EfnetB0 image transforms.
  """
  # Fix for wrong hash error from: https://github.com/pytorch/vision/issues/7744
  def get_state_dict(self, *args, **kwargs):
      kwargs.pop("check_hash")
      return load_state_dict_from_url(self.url, *args, **kwargs)
  WeightsEnum.get_state_dict = get_state_dict

  # Create EffNetB0 pretrained weights, transforms and model
  weights = EfficientNet_B0_Weights.DEFAULT
  transforms = weights.transforms()
  model = efficientnet_b0(weights=weights)

  # Freeze all layers in base model
  for param in model.features.parameters():
        param.requires_grad = False

  # Change the classifier head with random seed for reproducibility
  torch.manual_seed(seed)
  model.classifier = nn.Sequential(
      nn.Dropout(p=0.3),
      nn.Linear(in_features=1280, out_features=num_classes)
  )

  return model, transforms