"right-hurr-kick": [["down", "right", "z"], 0.025, 0.025, 1.75],
"low-heavy-kick": [["down", "ctrl"], 0.025, 0.025, 1],
"low-medium-kick": [["down", "z"], 0.025, 0.025, 0.75],
"low-light-kick": [["down", "x"], 0.025, 0.025, 0.75],
"low-heavy-punch": [["down", "shift"], 0.025, 0.025, 1],
"low-medium-punch": [["down", "v"], 0.025, 0.025, 1],
"low-light-punch": [["down", "c"], 0.025, 0.025, 1]
}
names_mov = list(movements.keys())
coord = (8, 95, 512, 475) #coordinates are different for every user...
env = Env.Enviroment(coord)
state = env.start_game(Open=True)
env.set_gamma(0.95)
Net = net.ActorCritic(state.shape, len(movements))
env.set_policy_net(Net)
file = r"\training_1\cp.ckpt"
def make_batches(env, batch=32, steps=1, gamma=0.95):
"""
Make batches of n_steps(bootstrapping)
Network expects float and Gray images, so we convert them
Returns: Initial_state,rewards,next_state,index of finished episodes,actions
"""
states = []
rewards = []
next_states = []
