qlearning algorithm to solve CartPole on openAI gym.

algorithm.py

def qlearning(env, policy, num_iter1, alpha, gamma):
    actions = policy.actions
    for i in xrange(len(policy.theta)):
        policy.theta[i] = 0.1

    for iter1 in xrange(num_iter1):
        s_f       = env.reset()
        a         = policy.epsilon_greedy(s_f)
        count     = 0
        t         = False

        while False == t and count < 10000:
            s_f1,r,t,i = env.step(a)
            qmax = policy.qfunc(s_f1,a) #random
            for a1 in policy.actions:
                pvalue = policy.qfunc(s_f1, a1)
                if qmax < pvalue:
                    qmax = pvalue;
            update(policy, s_f, a, r + gamma * qmax, alpha);

            s_f     = s_f1
            a       = policy.epsilon_greedy(s_f)
            count   += 1 
  
        if iter1%100 == 0:
            print "complete the %d epoches"%(iter1)

    return policy

Based on PEP8 you should avoid unnecessary whitespace.

Avoid more than one space around an assignment (or other) operator to align it with another.

Just a suggestion though.