浏览代码

Fix continuous curriosity

/develop-generalizationTraining-TrainerController
vincentpierre 6 年前
当前提交
1045b6e7
共有 1 个文件被更改,包括 1 次插入1 次删除
  1. 2
      ml-agents/mlagents/trainers/ppo/policy.py

2
ml-agents/mlagents/trainers/ppo/policy.py


feed_dict = {self.model.batch_size: len(next_info.vector_observations),
self.model.sequence_length: 1}
if self.use_continuous_act:
feed_dict[self.model.output] = next_info.previous_vector_actions
feed_dict[self.model.selected_actions] = next_info.previous_vector_actions
else:
feed_dict[self.model.action_holder] = next_info.previous_vector_actions
for i in range(self.model.vis_obs_size):

正在加载...
取消
保存