def __init__(self,
g_config = [512, 'leaky_relu', 1024, 'leaky_relu', 1024,
'leaky_relu', 256],
d_config = [128, 0.25, 'leaky_relu', 128, 0.25, 'leaky_relu', 1, 'sigmoid'],
batch_sz = 2048, n_epochs = 500):
super(ConditionalGAN, self).__init__()
self.G = nets.FullyConnectedUnits(g_config)
self.D = nets.FullyConnectedUnits(d_config)
self.batch_sz = batch_sz
self.ds = None
self.n_epochs = n_epochs
def test_single_task_training(file_path):
# load dataset
df = pd.read_csv(file_path)
y_tr = df.values[:, 1]
x_tr = df.values[:, -1]
lang_obj = lang.Lang(x_tr)
vocab_size = len(lang_obj.idx2ch)
x_tensor = lang.preprocessing(x_tr, lang_obj)
# define models
enc_f = nets.Encoder(vocab_size=vocab_size, reverse=False)
enc_b = nets.Encoder(vocab_size=vocab_size, reverse=True)
attention = nets.BidirectionalAttention(128)
fcuk = nets.FullyConnectedUnits([128, 'tanh', 0.25, 32, 'tanh', 0.10, 1])
# define the flow function
def flow(xs, models):
enc_f, enc_b, attention, fcuk = models
eo_f, h_f = enc_f(xs)
eo_b, h_b = enc_b(xs)
attention_weights = attention(eo_f, eo_b, h_f, h_b)
ys = fcuk(attention_weights)
return ys
box = nets.Box(flow=flow,
models=[enc_f, enc_b, attention, fcuk],
n_epochs=1,
batch_sz=32)
box.train(x_tensor, y_tr)
box.save_weights('box')
# save the dataset for later use
np.save('y_tr', y_tr)
np.save('x_tr', x_tr)
np.save('y_te', y_te)
np.save('x_te', x_te)
# create the language object and map it to strings
lang_obj = lang.Lang(list(x_tr) + list(x_te))
vocab_size = len(lang_obj.idx2ch) + 1
x_tr = lang.preprocessing(x_tr, lang_obj)
# define models
enc_f = nets.Encoder(vocab_size=vocab_size, batch_sz=BATCH_SZ, reverse=False)
enc_b = nets.Encoder(vocab_size=vocab_size, batch_sz=BATCH_SZ, reverse=True)
attention = nets.BidirectionalWideAttention(128)
fcuk = nets.FullyConnectedUnits(
[512, 'tanh', 0.30, 512, 'tanh', 0.30, 512, 'tanh', 0.25])
fcuk_props = nets.FullyConnectedUnits([9])
decoder = nets.AttentionDecoder(vocab_size=vocab_size)
# convert to tensor
x_tr = tf.convert_to_tensor(x_tr)
y_tr = tf.convert_to_tensor(y_tr)
# make them into a dataset object
ds = tf.data.Dataset.from_tensor_slices((x_tr, y_tr)).shuffle(y_tr.shape[0])
ds = ds.apply(tf.contrib.data.batch_and_drop_remainder(BATCH_SZ))
# get your favorite optimizer
optimizer = tf.train.AdamOptimizer()
# define models
enc_f = nets.GRUEncoder(vocab_size=vocab_size,
batch_sz=BATCH_SZ,
reverse=False,
enc_units=128)
enc_b = nets.GRUEncoder(vocab_size=vocab_size,
batch_sz=BATCH_SZ,
reverse=True,
enc_units=128)
conv_encoder = nets.ConvEncoder(
conv_units=[256, 512, 512],
# pool_sizes=[8, 8, 8, 8],
conv_kernel_sizes=[8, 12, 16],
fcs=[128, 0.2, 'elu', 512, 0.2, 'elu', 512])
fcuk = nets.FullyConnectedUnits([512, 'leaky_relu', 0.25, 512])
d_mean = nets.FullyConnectedUnits([32])
d_log_var = nets.FullyConnectedUnits([32])
fcuk_props = nets.FullyConnectedUnits([9])
fcuk_fp = nets.FullyConnectedUnits([167, 'sigmoid'])
decoder = nets.OneHotDecoder(vocab_size=vocab_size, dec_units=256)
bypass_v_f = nets.FullyConnectedUnits([1])
simple_decoder = nets.SimpleDecoder(vocab_size=vocab_size,
dec_units=1024,
batch_sz=BATCH_SZ)
# initialize
xs = tf.zeros([BATCH_SZ, 64], dtype=tf.int64)
eo_f, h_f = enc_f(xs)
eo_b, h_b = enc_b(xs)
x_tr = np.load('x_tr.npy')
x_te = np.load('x_te.npy')
y_tr = np.load('y_tr.npy')
y_te = np.load('y_te.npy')
# create the language object and map it to strings
lang_obj = lang.Lang(list(x_tr) + list(x_te))
vocab_size = len(lang_obj.idx2ch) + 1
x_tr = lang.preprocessing(x_tr, lang_obj)
# define models
enc_f = nets.Encoder(vocab_size=vocab_size, batch_sz = BATCH_SZ, reverse=False)
enc_b = nets.Encoder(vocab_size=vocab_size, batch_sz = BATCH_SZ, reverse=True)
attention = nets.BidirectionalWideAttention(128)
fcuk = nets.FullyConnectedUnits([128, 'tanh', 0.30, 1024, 'tanh', 0.30, 1024,
'tanh', 0.25, 256, 'tanh', 0.10])
fcuk_props = nets.FullyConnectedUnits([9])
decoder = nets.AttentionDecoder(vocab_size=vocab_size)
# convert to tensor
x_tr = tf.convert_to_tensor(x_tr)
y_tr = tf.convert_to_tensor(y_tr)
# initialize
xs = tf.zeros((BATCH_SZ, 64))
eo_f, h_f = enc_f(xs)
eo_b, h_b = enc_b(xs)
attention_weights = attention(eo_f, eo_b, h_f, h_b)
attention_weights = fcuk(attention_weights)
ys_hat = fcuk_props(attention_weights)
dec_input = tf.expand_dims([lang_obj.ch2idx['G']] * BATCH_SZ, 1)