mattbui / dd_qnet Goto Github PK

Hi, when I train the laser model by running gazebo_turtlebot_maze_color_laser_only.py, I have an error.
In utils.py, "parser.add_argument("--num_pretrain_step", type=int, default=10000)". This definition shows that the first 10,000 steps are all randomly selecting actions. When it exceeds 10,000, it starts to select actions from the network.
But after the steps exceed 10000, I encountered the following error in terminal:

`Done epoch in 10 steps, 10 random steps, Total reward: -191, Total step: 4981, Average loss: 0.0
Setting target 1th
Traceback (most recent call last):

File "laser_learn.py", line 48, in
action = qnet.get_actions(state.reshape(1, -1))[0]
File "/home/nuc/gym-gazebo/dd_qnet/src/laser_model.py", line 55, in get_actions
qvalues = self.get_qvalues(states)
File "/home/nuc/gym-gazebo/dd_qnet/src/laser_model.py", line 47, in get_qvalues
predicted = self.model.predict(states)
File "/home/nuc/.local/lib/python3.5/site-packages/keras/engine/training.py", line 1152, in predict
x, _, _ = self._standardize_user_data(x)
File "/home/nuc/.local/lib/python3.5/site-packages/keras/engine/training.py", line 754, in _standardize_user_data
exception_prefix='input')
File "/home/nuc/.local/lib/python3.5/site-packages/keras/engine/training_utils.py", line 136, in standardize_input_data
str(data_shape))

ValueError: Error when checking input: expected input_1 to have shape (100,) but got array with shape (1,)`

Looking forward to your reply, thank you.

hello, May I ask, can pre-trained models be provided? The link you gave is invalid.