parser.add_argument("--batch_size", type=int, default=32) parser.add_argument("--n_imgs", type=int, default=16) parser.add_argument("--seed", default=1, type=int, help="Random seed") args = parser.parse_args() if __name__ == "__main__": torch.manual_seed(args.seed) np.random.seed(args.seed) agent = Agent(args.n_imgs) agent.init_from_save(filename=f'{args.output_dir}/{args.model_name}.pkl') agent.prep_eval() env = Env(args.n_imgs) env.batch_size = args.batch_size env.load_labels(data_path='data/train.txt') env.load_video(video_path='data/train.mp4') env.shuffle_data() env.prep_eval() criterion = nn.MSELoss() test_records = EvalRecords() for i_ep in range(1000): state, labels = env.get_data() torch_state = Variable(torch.from_numpy(state)) torch_labels = Variable(torch.from_numpy(labels)) # forward outputs = agent.predict(torch_state)