def __init__(self):
super(VAERNN, self).__init__()
self.z_size = 32
self.kl_tolerance = 0.5
self.vae = VAE()
self.rnn = RNN()
self.vae.train()
self.rnn.train()
self.init_()
self.is_cuda = False
def __init__(self):
super(VAERNN, self).__init__()
self.z_size = 32
self.kl_tolerance = 0.5
self.vae = VAE()
self.rnn = RNN()
self.vae.load_state_dict(
torch.load(vae_model_path,
map_location=lambda storage, loc: storage))
self.rnn.load_state_dict(
torch.load(rnn_model_path,
map_location=lambda storage, loc: storage))
self.vae.train()
self.rnn.train()
self.init_()
self.is_cuda = False
import torch.multiprocessing as mp
import torch.optim as optim
from this_util import *
from vae import VAE
from rnn_me import RNN
from policy import Policy
vae_model = VAE()
vae_model.load_state_dict(
torch.load(vae_model_path, map_location=lambda storage, loc: storage)
) #, map_location=lambda storage, loc: storage)
vae_model.eval()
rnn_model = RNN()
rnn_model.load_state_dict(
torch.load(rnn_model_path, map_location=lambda storage, loc: storage))
rnn_model.eval()
def wow(state, h, policy, vae_model, rnn_model):
state = tensor_state(state)
z = vae_model(state)
h = h.squeeze(0)
z_h = torch.cat((z, h), dim=1)
a = policy(z_h)
one = one_hot(a)
one = torch.from_numpy(one)
one = one.unsqueeze(0)
