Gym: Setting random seed to reproduce the same results

[DollarAkshay]

I tried setting the seed by using

random.seed(1995)

But I do not get the same results. I looks like every game environment initializes its own unique seed. If this is the case how would I go about generating the same results every time ? In this case im trying to do this for LunarLander-v2

[DollarAkshay]

Edit : I also just tried

observation = env.reset()
env.seed(1995)

It is still giving different results. even after setting the seed after every reset()

The code in question is located here : https://github.com/DollarAkshay/Python-Programs/blob/master/Artificial%20Intelligence/OpenAI_LunarLander_v2.py

[jietang]

env.seed() will seed the environment randomness. If your agent also uses randomness, you will need to seed that separately (i.e. using random.seed()) to get repeatable results

ex://

>>> import gym
>>> env = gym.make('LunarLander-v2')
[2016-12-21 10:38:47,791] Making new env: LunarLander-v2

>>> env.seed(0)
[0L]
>>> env.reset()
array([ -5.91564178e-04,   9.42304904e-01,  -5.99357188e-02,
         1.12770955e-01,   6.92289264e-04,   1.35763153e-02,
         0.00000000e+00,   0.00000000e+00])
>>> env.step(0)
(array([-0.00118332,  0.94361125, -0.05985444,  0.08708715,  0.00136317,
         0.01341959,  0.        ,  0.        ]), 2.0058447786388598, False, {})

>>> env.seed(0)
[0L]
>>> env.reset()
array([ -5.91564178e-04,   9.42304904e-01,  -5.99357188e-02,
         1.12770955e-01,   6.92289264e-04,   1.35763153e-02,
         0.00000000e+00,   0.00000000e+00])
>>> env.step(0)
(array([-0.00118332,  0.94361125, -0.05985444,  0.08708715,  0.00136317,
         0.01341959,  0.        ,  0.        ]), 2.0058447786388598, False, {})