search

voiler / unreal

Reinforcement learning with unsupervised auxiliary tasks
23 stars 4 forks source link

LSTM State is missing in generating the action #2

Open Shengyu-Feng opened 5 years ago

Shengyu-Feng commented 5 years ago

I think when actor is generating the next state, it fails to use the initial state from last cell, as in this line trainer.py

QimingLiuSJTU commented 3 years ago

Agree