import torch
from fastai.vision.all import *

def load_model(model_path):
    # Load the model weights from the .pth file
    state_dict = torch.load(model_path)

    # Define the model architecture
    model = resnet34(num_classes=2)

    # Load the model weights into the architecture
    model.load_state_dict(state_dict)

    # Define the data loaders
    dls = ImageDataLoaders.from_folder(path, train='train', valid='valid')

    # Define the Learner object
    learn = Learner(dls, model, metrics=accuracy)

    return learn

# Load the model from the .pth file and create the necessary objects
#learn = load_model('my_model.pth')

# Use the model for inference or further training