model.py

import torch

from pathlib import Path
from torch import nn
from torchvision.models import efficientnet_b2, EfficientNet_B2_Weights
from torchvision.models._api import WeightsEnum
from torch.hub import load_state_dict_from_url
from typing import Optional, Tuple

DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'

# A RuntimeError appeared in HuggingFace when the application tried to load the
# weights of the model. The following link is to the source of the fix now
# implemented.
# https://github.com/pytorch/vision/issues/7744#issuecomment-1757321451


def get_state_dict(self, *args, **kwargs):
    """
        Override intented to fix a bug while loading the state_dict
        from the internet.
    """
    kwargs.pop("check_hash")
    return load_state_dict_from_url(self.url, *args, **kwargs)


WeightsEnum.get_state_dict = get_state_dict


def create_effnetb2_model(
    num_classes: int,
    seed: Optional[int] = 42,
    load_st_dict: Optional[bool] = False
) -> Tuple[nn.Module, nn.Module]:
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)
    weights = EfficientNet_B2_Weights.DEFAULT
    transforms = weights.transforms()
    model = efficientnet_b2(weights=weights)

    model.classifier = nn.Sequential(
      nn.Dropout(p=0.3, inplace=True),
      nn.Linear(in_features=1408, out_features=num_classes, bias=True)
    ).to(DEVICE)
    if load_st_dict:
        st_dict = Path('model.pth')
        model.load_state_dict(torch.load(st_dict, map_location=DEVICE))
    for param in model.parameters():
        param.requires_grad = False

    return model.to(DEVICE), transforms