There are two typos on line 70 and 168 of a3c.py, which result in an input error, I think,
split_5 = tflearn.fully_connected(inputs[:, **4:5**, -1], 128, activation='relu')
should be
split_5 = tflearn.fully_connected(inputs[:, **5:6**, -1], 128, activation='relu')
Hi, Hongzi,
There are two typos on line 70 and 168 of a3c.py, which result in an input error, I think,
split_5 = tflearn.fully_connected(inputs[:, **4:5**, -1], 128, activation='relu')
should besplit_5 = tflearn.fully_connected(inputs[:, **5:6**, -1], 128, activation='relu')