class World(Task):
+ def save_image(self, input, result_dir, filename, logger):
+ img = world.sample2img(self.train_input.to("cpu"), self.height, self.width)
+ image_name = os.path.join(result_dir, filename)
+ torchvision.utils.save_image(img.float() / 255.0, image_name, nrow=8, padding=2)
+ logger(f"wrote {image_name}")
+
def __init__(
self,
nb_train_samples,
nb_test_samples,
batch_size,
+ result_dir=None,
logger=None,
device=torch.device("cpu"),
):
self.nb_codes = max(self.train_input.max(), self.test_input.max()) + 1
+ if result_dir is not None:
+ self.save_image(
+ self.train_input[:96], result_dir, f"world_train.png", logger
+ )
+
def batches(self, split="train", nb_to_use=-1, desc=None):
assert split in {"train", "test"}
input = self.train_input if split == "train" else self.test_input
##############################
- input, ar_mask = self.test_input[:64], self.test_ar_mask[:64]
+ input, ar_mask = self.test_input[:96], self.test_ar_mask[:96]
result = input.clone() * (1 - ar_mask)
masked_inplace_autoregression(
device=self.device,
)
- img = world.sample2img(result.to("cpu"), self.height, self.width)
- image_name = os.path.join(result_dir, f"world_result_{n_epoch:04d}.png")
- torchvision.utils.save_image(img.float() / 255.0, image_name, nrow=8, padding=2)
- logger(f"wrote {image_name}")
+ self.save_image(result, result_dir, f"world_result_{n_epoch:04d}.png", logger)
+
+ def store_new_problems(self, new_problems):
+ nb_current = self.train_input.size(0)
+ nb_new = new_problems.size(0)
+ if nb_new >= nb_current:
+ self.train_input[...] = new_problems[:nb_current]
+ else:
+ nb_kept = nb_current - nb_new
+ self.train_input[:nb_kept] = self.train_input[-nb_kept:].clone()
+ self.train_input[nb_kept:] = new_problems
def create_new_problems(self, n_epoch, result_dir, logger, nb, model, nb_runs):
new_problems = torch.empty(
device=self.device,
)
- img = world.sample2img(new_problems[:64].to("cpu"), self.height, self.width)
- image_name = os.path.join(result_dir, f"world_new_{n_epoch:04d}.png")
- torchvision.utils.save_image(img.float() / 255.0, image_name, nrow=8, padding=2)
- logger(f"wrote {image_name}")
-
nb_correct = torch.empty(nb, device=self.device, dtype=torch.int64)
for n in tqdm.tqdm(