changjingzhi / rl_dqn_stock Goto Github PK

Hello, I am able to have partial success in the first round of the game when I run it on my dataset using your test_DQN.py.

barpos: 3397
sell:not enough stock
store_transition index: 97
episode 0 profits -258917.04 avg profits -258739.44 epsilon 0.03
barpos: 3398
sell:not enough stock
store_transition index: 98
episode 0 profits -258870.77 avg profits -258690.19 epsilon 0.03
barpos: 3399
sell:not enough stock
store_transition index: 99
episode 0 profits -258558.04 avg profits -258644.00 epsilon 0.03

It was a poor result, but it was only the first round of play.
However, after this, it suddenly reported an error，

barpos: 3399
sell:not enough stock
store_transition index: 99
episode 0 profits -258558.04 avg profits -258644.00 epsilon 0.03
barpos: 0
Traceback (most recent call last):
File "D:\py_code\DQN_base-main\stock\DQN.py", line 251, in
action = agent.choose_action(observation)
File "D:\py_code\DQN_base-main\stock\DQN.py", line 186, in choose_action
state = T.tensor(observation).to(self.Q_eval.device)
ValueError: too many dimensions 'str'

I don't understand why this is so, first of all, my data total is 3,401.It looks like a mistake was made at the start of the second round.
I hope to get your answer.