import torch

class M5(torch.nn.Module):
    def __init__(self, num_classes=2):  # Ensure it matches dataset labels (chainsaw/environment)
        super(M5, self).__init__()
        self.conv1 = torch.nn.Conv1d(in_channels=1, out_channels=32, kernel_size=80, stride=4)
        self.bn1 = torch.nn.BatchNorm1d(32)
        self.conv2 = torch.nn.Conv1d(in_channels=32, out_channels=64, kernel_size=3)
        self.bn2 = torch.nn.BatchNorm1d(64)
        self.conv3 = torch.nn.Conv1d(in_channels=64, out_channels=128, kernel_size=3)
        self.bn3 = torch.nn.BatchNorm1d(128)
        self.conv4 = torch.nn.Conv1d(in_channels=128, out_channels=256, kernel_size=3)
        self.bn4 = torch.nn.BatchNorm1d(256)
        self.fc1 = torch.nn.Linear(256, num_classes)

    def forward(self, x):
        x = torch.nn.functional.relu(self.bn1(self.conv1(x)))
        x = torch.nn.functional.max_pool1d(x, 4)
        x = torch.nn.functional.relu(self.bn2(self.conv2(x)))
        x = torch.nn.functional.max_pool1d(x, 4)
        x = torch.nn.functional.relu(self.bn3(self.conv3(x)))
        x = torch.nn.functional.max_pool1d(x, 4)
        x = torch.nn.functional.relu(self.bn4(self.conv4(x)))
        x = torch.nn.functional.max_pool1d(x, 4)
        x = torch.mean(x, dim=2)
        x = self.fc1(x)
        return x

def load_model(model_path, num_classes=2):
    """
    Load trained M5 model.
    """
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model = M5(num_classes=num_classes).to(device)
    model.load_state_dict(torch.load(model_path, map_location=device))
    model.eval()  # Set model to evaluation mode
    return model, device

if __name__ == "__main__":
    model, device = load_model("quantized_teacher_m5_static.pth")
    print("✅ Model successfully loaded!")