##############################
# maze options
-parser.add_argument("--world_height", type=int, default=13)
+parser.add_argument("--maze_height", type=int, default=13)
-parser.add_argument("--world_width", type=int, default=21)
+parser.add_argument("--maze_width", type=int, default=21)
-parser.add_argument("--world_nb_walls", type=int, default=15)
+parser.add_argument("--maze_nb_walls", type=int, default=15)
######################################################################
result = input.clone()
ar_mask = result.new_zeros(result.size())
ar_mask[:, self.height * self.width :] = 1
+ result *= 1-ar_mask
masked_inplace_autoregression(model, self.batch_size, result, ar_mask)
mazes, paths = self.seq2map(result)
nb_correct += maze.path_correctness(mazes, paths).long().sum()
input = self.test_input[:32]
result = input.clone()
ar_mask = result.new_zeros(result.size())
-
ar_mask[:, self.height * self.width :] = 1
+ result *= 1-ar_mask
masked_inplace_autoregression(model, self.batch_size, result, ar_mask)
mazes, paths = self.seq2map(input)
_, predicted_paths = self.seq2map(result)
- maze.save_image(f"result_{n_epoch:04d}.png", mazes, paths, predicted_paths)
+ maze.save_image(
+ f"result_{n_epoch:04d}.png",
+ mazes,
+ paths,
+ predicted_paths,
+ maze.path_correctness(mazes, predicted_paths),
+ )
model.train(t)
nb_train_samples=args.nb_train_samples,
nb_test_samples=args.nb_test_samples,
batch_size=args.batch_size,
- height=args.world_height,
- width=args.world_width,
- nb_walls=args.world_nb_walls,
+ height=args.maze_height,
+ width=args.maze_width,
+ nb_walls=args.maze_nb_walls,
device=device,
)