from models.text2video_model import Text2VideoModel
from training.trainer import Text2VideoTrainer
from config.model_config import CONFIG
import torch

def main():
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    
    model = Text2VideoModel(
        vocab_size=CONFIG['vocab_size'],
        embed_dim=CONFIG['embed_dim'],
        latent_dim=CONFIG['latent_dim'],
        num_frames=CONFIG['num_frames'],
        frame_size=CONFIG['frame_size']
    ).to(device)
    
    optimizer = torch.optim.Adam(model.parameters(), lr=CONFIG['learning_rate'])
    trainer = Text2VideoTrainer(model, optimizer, device)
    
    # Add your data loading and training loop here

if __name__ == '__main__':
    main()
    
class Text2VideoTrainer:
    def __init__(self, model, optimizer, device):
        self.model = model
        self.optimizer = optimizer
        self.device = device
        
    def train_step(self, text_batch, video_batch):
        self.optimizer.zero_grad()
        
        generated_video = self.model(text_batch)
        loss = F.mse_loss(generated_video, video_batch)
        
        loss.backward()
        self.optimizer.step()
        
        return loss.item()