def step(self, action):
self.canvas = (decode(action, self.canvas.float() / 255) * 255).byte()
self.stepnum += 1
ob = self.observation()
done = (self.stepnum == self.max_step)
reward = self.cal_reward() # np.array([0.] * self.batch_size)
return ob.detach(), reward, np.array([done] * self.batch_size), None
def step(self, action, episode_num):
if self.use_multiple_renderers:
self.canvas = (decode_multiple_renderers(
action,
self.canvas.float() / 255, episode_num) * 255).byte()
else:
self.canvas = (decode(action,
self.canvas.float() / 255) * 255).byte()
self.stepnum += 1
ob, mask = self.observation()
done = (self.stepnum == self.max_step)
reward = self.cal_reward() # np.array([0.] * self.batch_size)
return ob.detach(), reward, np.array([done] *
self.batch_size), None, mask
def step(self, action):
#if torch.is_tensor(self.prev_action):
# strokeの間を繋ぐ
#new_action = [
# self.prev_action[:,4:6],
# self.prev_action[:,4:6],
# action[:,:2],
# self.prev_action[:,7:8],
# action[:,6:7],
# action[:,8:13]
#]
#connect_action = torch.cat(new_action, dim=1)
#self.canvas = (decode(connect_action, self.canvas.float() / 255) * 255).byte()
self.canvas = (decode(action, self.canvas.float() / 255) * 255).byte()
self.stepnum += 1
self.prev_action = action
ob = self.observation()
done = (self.stepnum == self.max_step)
reward = self.cal_reward() # np.array([0.] * self.batch_size)
return ob.detach(), reward, np.array([done] * self.batch_size), None
