What the requirements to run the source code ?

[developeralgo8888]

i am not sure what is needed in terms of Tensorflow and Keras versions, there is no clarity at all

Tensorflow 2.3+ which already contains keras Keras_rl ????? ( this library is mostly outdated ) instead tf-agents is much better

Can you be specific on what are the requirements in terms of library versions needed ?

[renan-r-santos]

Hi, I managed to run the source code using Miniconda and the commands below. By the way, apparently this code doesn't include layer 1 code (CNN training mentioned in the article).

conda create --name tensorflow python=3.7
conda install tensorflow-gpu=1.15.0
conda activate tensorflow
conda install keras pandas=0.23.4 matplotlib
pip install keras-rl gym keras-radam

EDIT: you need to enforce tensorflow version. The code is not for TF 2

EDIT 2: you need an older version of pandas

EDIT 3: found a bug on main.py. Argv index is wrong in line 78 and onwards. I should be 3 instead of 4

if(sys.argv[3]=="relu"):
    model.add(Dense(35,activation='relu'))    
if(sys.argv[3]=="sigmoid"):
    model.add(Dense(35,activation='sigmoid'))    
if(sys.argv[3]=="linear"):
    model.add(Dense(35,activation='linear'))
if(sys.argv[3]=="tanh"):
    model.add(Dense(35,activation='tanh'))
if(sys.argv[3]=="selu"):
    model.add(Dense(35,activation='selu'))

EDIT 4: I had to comment lines 108 and 109 of the file dqn.py (keras-rl package):

# if hasattr(model.output, '__len__') and len(model.output) > 1:
#     raise ValueError('Model "{}" has more than one output. DQN expects a model that has a single output.'.format(model))

[wac81]

Hi, I managed to run the source code using Miniconda and the commands below. By the way, apparently this code doesn't include layer 1 code (CNN training mentioned in the article).

conda create --name tensorflow python=3.7
conda install tensorflow-gpu=1.15.0
conda activate tensorflow
conda install keras pandas=0.23.4 matplotlib
pip install keras-rl gym keras-radam

EDIT: you need to enforce tensorflow version. The code is not for TF 2

EDIT 2: you need an older version of pandas

EDIT 3: found a bug on main.py. Argv index is wrong in line 78 and onwards. I should be 3 instead of 4

if(sys.argv[3]=="relu"):
    model.add(Dense(35,activation='relu'))    
if(sys.argv[3]=="sigmoid"):
    model.add(Dense(35,activation='sigmoid'))    
if(sys.argv[3]=="linear"):
    model.add(Dense(35,activation='linear'))
if(sys.argv[3]=="tanh"):
    model.add(Dense(35,activation='tanh'))
if(sys.argv[3]=="selu"):
    model.add(Dense(35,activation='selu'))

EDIT 4: I had to comment lines 108 and 109 of the file dqn.py (keras-rl package):

# if hasattr(model.output, '__len__') and len(model.output) > 1:
#     raise ValueError('Model "{}" has more than one output. DQN expects a model that has a single output.'.format(model))

thank you ， good suggestion