model.py

import torch.nn as nn
import torchvision.models as models


class Musheff(nn.Module):
    def __init__(self, config):
        super().__init__()

        # Extract parameters from config
        num_classes = config["num_classes"]
        dropout_rate = config["dropout_rate"]

        # Load default weights from base model
        weights = models.EfficientNet_B3_Weights.DEFAULT

        # Load base model
        self.model = models.efficientnet_b3(weights=weights)

        # Modify classifier head
        in_features = self.model.classifier[1].in_features
        self.model.classifier = nn.Sequential(
            nn.Dropout(p=dropout_rate, inplace=True),
            nn.Linear(in_features, num_classes),
        )

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