model.py

import torch
import torchvision

from torch import nn


def create_model(num_classes=101, seed=42):
    
    weights = torchvision.models.ConvNeXt_Tiny_Weights.DEFAULT # .DEFAULT = best available weights on ImageNet
    transforms = weights.transforms()
    model = torchvision.models.convnext_tiny(weights=weights)

    # Sequential (features)
    for param in model.features.parameters(): 
        param.requires_grad = False # "requires" "grad"ient-descent

    # Sequential (classifier)
    torch.manual_seed(seed)
    model.classifier[-1] = nn.Linear(in_features=model.classifier[-1].in_features, 
                                        out_features=num_classes)
    
    return model, transforms