token_background = 0
first_bird_token = 1
nb_bird_tokens = colors.size(0) - 1
- token_forward = first_bird_token + nb_bird_tokens
- token_backward = token_forward + 1
token2char = (
"_" + "".join([chr(ord("A") + n) for n in range(len(colors) - 1)]) + "><"
)
- def __init__(self, height=6, width=8, nb_birds=3, speed=1, nb_iterations=4):
+ def nb_token_values(self):
+ return len(self.colors)
+
+ def __init__(
+ self,
+ height=6,
+ width=8,
+ nb_birds=3,
+ speed=2,
+ nb_iterations=2,
+ avoid_collision=True,
+ ):
self.height = height
self.width = width
self.nb_birds = nb_birds
self.speed = speed
self.nb_iterations = nb_iterations
+ self.avoid_collision = avoid_collision
- def direction_tokens(self):
- return self.token_forward, self.token_backward
-
- def generate_seq(self, nb, return_frame_sequences=False):
+ def generate_frame_sequences(self, nb):
frame_sequences = []
for _ in tqdm.tqdm(range(nb), dynamic_ncols=True, desc="world generation"):
- result = torch.zeros(
- self.nb_iterations, self.height, self.width, dtype=torch.int64
- )
-
i, j, vi, vj = (
torch.empty(self.nb_birds, dtype=torch.int64),
torch.empty(self.nb_birds, dtype=torch.int64),
torch.empty(self.nb_birds, dtype=torch.int64),
)
+ def collision_okay():
+ if not self.avoid_collision:
+ return True
+
+ count = torch.zeros(self.height, self.width, dtype=torch.int64)
+
+ for n in range(self.nb_birds):
+ count[i[n], j[n]] += 1
+ count[i[n] - vi[n], j[n]] += 1
+ count[i[n], j[n] - vj[n]] += 1
+
+ return count.max() <= 1
+
col = (
torch.randperm(self.colors.size(0) - 1)[: self.nb_birds].sort().values
+ 1
)
- for n in range(self.nb_birds):
+ while True:
while True:
- i[n] = torch.randint(self.height, (1,))
- j[n] = torch.randint(self.width, (1,))
- vm = torch.randint(4, (1,))
- vi[n], vj[n] = (vm % 2) * 2 - 1, (vm // 2) * 2 - 1
- if (
- i[n] - vi[n] >= 0
- and i[n] - vi[n] < self.height
- and j[n] - vj[n] >= 0
- and j[n] - vj[n] < self.width
- ):
+ for n in range(self.nb_birds):
+ while True:
+ i[n] = torch.randint(self.height, (1,))
+ j[n] = torch.randint(self.width, (1,))
+ vm = torch.randint(4, (1,))
+ vi[n], vj[n] = (vm % 2) * 2 - 1, (vm // 2) * 2 - 1
+ if (
+ i[n] - vi[n] >= 0
+ and i[n] - vi[n] < self.height
+ and j[n] - vj[n] >= 0
+ and j[n] - vj[n] < self.width
+ ):
+ break
+
+ if collision_okay():
break
- for l in range(self.nb_iterations):
- for n in range(self.nb_birds):
- c = col[n]
- result[l, i[n], j[n]] = c
- result[l, i[n] - vi[n], j[n]] = c
- result[l, i[n], j[n] - vj[n]] = c
+ result = torch.zeros(
+ self.nb_iterations * self.speed,
+ self.height,
+ self.width,
+ dtype=torch.int64,
+ )
- if (i[n] == 0 and vi[n] == -1) or (
- i[n] == self.height - 1 and vi[n] == 1
- ):
- vi[n] = -vi[n]
+ fine = torch.empty(self.nb_iterations * self.speed)
- if (j[n] == 0 and vj[n] == -1) or (
- j[n] == self.width - 1 and vj[n] == 1
- ):
- vj[n] = -vj[n]
+ t_to_keep = (
+ torch.arange(self.nb_iterations, device=result.device) * self.speed
+ )
- i[n] += vi[n]
- j[n] += vj[n]
+ for l in range(self.nb_iterations * self.speed):
+ fine[l] = collision_okay()
+ for n in range(self.nb_birds):
+ c = col[n]
+ result[l, i[n], j[n]] = c
+ result[l, i[n] - vi[n], j[n]] = c
+ result[l, i[n], j[n] - vj[n]] = c
- frame_sequences.append(result)
+ if (i[n] == 0 and vi[n] == -1) or (
+ i[n] == self.height - 1 and vi[n] == 1
+ ):
+ vi[n] = -vi[n]
- if return_frame_sequences:
- return frame_sequences
+ if (j[n] == 0 and vj[n] == -1) or (
+ j[n] == self.width - 1 and vj[n] == 1
+ ):
+ vj[n] = -vj[n]
- # Randomize the time direction, annd convert to token
- # sequences with the time direction tokens added
+ i[n] += vi[n]
+ j[n] += vj[n]
- result = []
+ result = result[t_to_keep]
+ fine = fine[t_to_keep]
- for frame_sequence in frame_sequences:
- a = []
- if torch.rand(1) < 0.5:
- for frame in frame_sequence:
- if len(a) > 0:
- a.append(torch.tensor([self.token_forward]))
- a.append(frame.flatten())
- else:
- for frame in reversed(frame_sequence):
- if len(a) > 0:
- a.append(torch.tensor([self.token_backward]))
- a.append(frame.flatten())
+ if fine[-1]:
+ break
+
+ frame_sequences.append(result)
- result.append(torch.cat(a, dim=0)[None, :])
+ return frame_sequences
- return torch.cat(result, dim=0)
+ ######################################################################
+
+ def generate_prompts_and_answers(self, nb):
+ frame_sequences = self.generate_frame_sequences(nb)
+ frame_sequences = torch.cat([x[None] for x in frame_sequences], dim=0)
+ prompts = frame_sequences[:, : frame_sequences.size(1) // 2].flatten(1)
+ answers = frame_sequences[:, frame_sequences.size(1) // 2 :].flatten(1)
+ return prompts, answers
######################################################################
return x
- def seq2img(self, seq, scale=15):
- all = [
- self.frame2img(
- seq[:, : self.height * self.width].reshape(-1, self.height, self.width),
- scale,
+ def seq2str(self, seq):
+ result = []
+ for s in seq:
+ result.append("".join([self.token2char[v] for v in s]))
+ return result
+
+ def save_image(
+ self,
+ result_dir,
+ filename,
+ prompts,
+ answers,
+ predicted_prompts=None,
+ predicted_answers=None,
+ ):
+ if predicted_prompts is None:
+ predicted_prompts = 255
+
+ if predicted_answers is None:
+ predicted_answers = 255
+
+ def add_frame(x, c, margin):
+ y = x.new_full(
+ (x.size(0), x.size(1), x.size(2) + 2 * margin, x.size(3) + 2 * margin),
+ 0,
)
- ]
+ if type(c) is int:
+ y[...] = c
+ else:
+ c = c.long()[:, None]
+ c = c * torch.tensor([192, 192, 192], device=c.device) + (
+ 1 - c
+ ) * torch.tensor([255, 255, 255], device=c.device)
+ y[...] = c[:, :, None, None]
+ y[:, :, margin:-margin, margin:-margin] = x
+ return y
- separator = torch.full((seq.size(0), 3, self.height * scale - 1, 1), 0)
+ margin = 4
- t = self.height * self.width
+ img_prompts = add_frame(self.frame2img(prompts.to("cpu")), 0, 1)
+ img_answers = add_frame(self.frame2img(answers.to("cpu")), 0, 1)
- while t < seq.size(1):
- direction_tokens = seq[:, t]
- t += 1
+ # img_prompts = add_frame(img_prompts, 255, margin)
+ # img_answers = add_frame(img_answers, 255, margin)
- direction_images = self.colors[
- torch.full(
- (direction_tokens.size(0), self.height * scale - 1, scale), 0
- )
- ].permute(0, 3, 1, 2)
-
- for n in range(direction_tokens.size(0)):
- if direction_tokens[n] == self.token_forward:
- for k in range(scale):
- for l in [0, 1]:
- direction_images[
- n,
- :,
- (self.height * scale) // 2 - scale // 2 + k - l,
- 3 + scale // 2 - abs(k - scale // 2),
- ] = 0
- elif direction_tokens[n] == self.token_backward:
- for k in range(scale):
- for l in [0, 1]:
- direction_images[
- n,
- :,
- (self.height * scale) // 2 - scale // 2 + k - l,
- 3 + abs(k - scale // 2),
- ] = 0
- else:
- for k in range(2, scale - 2):
- for l in [0, 1]:
- direction_images[
- n,
- :,
- (self.height * scale) // 2 - scale // 2 + k - l,
- k,
- ] = 0
- direction_images[
- n,
- :,
- (self.height * scale) // 2 - scale // 2 + k - l,
- scale - 1 - k,
- ] = 0
-
- all += [
- separator,
- direction_images,
- separator,
- self.frame2img(
- seq[:, t : t + self.height * self.width].reshape(
- -1, self.height, self.width
- ),
- scale,
- ),
- ]
-
- t += self.height * self.width
-
- return torch.cat(all, dim=3)
+ img_prompts = add_frame(img_prompts, predicted_prompts, margin)
+ img_answers = add_frame(img_answers, predicted_answers, margin)
- def seq2str(self, seq):
- result = []
- for s in seq:
- result.append("".join([self.token2char[v] for v in s]))
- return result
+ separator = img_prompts.new_full(
+ (img_prompts.size(0), img_prompts.size(1), img_prompts.size(2), margin), 255
+ )
- def save_image(self, input, result_dir, filename):
- img = self.seq2img(input.to("cpu"))
- image_name = os.path.join(result_dir, filename)
- torchvision.utils.save_image(img.float() / 255.0, image_name, nrow=6, padding=4)
+ img = torch.cat([img_prompts, img_answers], dim=3)
- def save_quizzes(self, input, result_dir, filename_prefix):
- self.save_image(input, result_dir, filename_prefix + ".png")
+ image_name = os.path.join(result_dir, filename)
+ torchvision.utils.save_image(
+ img.float() / 255.0, image_name, nrow=6, padding=margin * 2, pad_value=1.0
+ )
+
+ def save_quizzes(
+ self,
+ result_dir,
+ filename_prefix,
+ prompts,
+ answers,
+ predicted_prompts=None,
+ predicted_answers=None,
+ ):
+ self.save_image(
+ result_dir,
+ filename_prefix + ".png",
+ prompts,
+ answers,
+ predicted_prompts,
+ predicted_answers,
+ )
######################################################################
if __name__ == "__main__":
import time
- sky = Sky(height=6, width=8, speed=1, nb_iterations=4)
+ sky = Sky(height=6, width=8, speed=4, nb_iterations=2)
+
+ prompts, answers = sky.generate_prompts_and_answers(4)
+
+ predicted_prompts = torch.rand(prompts.size(0)) < 0.5
+ predicted_answers = torch.rand(answers.size(0)) < 0.5
- start_time = time.perf_counter()
- seq = sky.generate_seq(nb=64)
- delay = time.perf_counter() - start_time
- print(f"{seq.size(0)/delay:02f} seq/s")
+ sky.save_quizzes(
+ "/tmp", "test", prompts, answers, predicted_prompts, predicted_answers
+ )
+
+ # start_time = time.perf_counter()
+ # token_sequences = sky.generate_token_sequences(nb=64)
+ # delay = time.perf_counter() - start_time
+ # print(f"{token_sequences.size(0)/delay:02f} seq/s")
# print(sky.seq2str(seq[:4]))
# m = (torch.rand(seq.size()) < 0.05).long()
# seq = (1 - m) * seq + m * 23
- print(seq.size())
- img = sky.seq2img(seq)
- print(img.size())
+ # print(seq.size())
+ # img = sky.seq2img(token_sequences)
+ # print(img.size())
- torchvision.utils.save_image(
- img.float() / 255.0, "/tmp/world.png", nrow=6, padding=6, pad_value=0
- )
+ # torchvision.utils.save_image(
+ # img.float() / 255.0, "/tmp/world.png", nrow=6, padding=6, pad_value=0
+ # )