def ner_accuracy(tensor, opt):
r"""Returns accuracy of predictions.
Args:
tensor: A `Tensor`. Probability distributions or unscaled prediction scores.
opt:
target: A 'Tensor`. Labels.
Returns:
A `Tensor` of the same shape as `tensor`. Each value will be 1 if correct else 0.
For example,
```
tensor = [[20.1, 18, -4.2], [0.04, 21.1, 31.3]]
target = [[0, 1]]
tensor.sg_accuracy(target=target) => [[ 1. 0.]]
```
"""
assert opt.target is not None, 'target is mandatory.'
opt += tf.sg_opt(k=1)
# # calc accuracy
out = tf.identity(tf.equal(tensor.sg_argmax() + 1,
tf.cast(opt.target, tf.int64)).sg_float(),
name='acc')
# out = tf.identity(tf.nn.in_top_k(tensor, opt.target, opt.k).sg_float(), name='acc')
# masking padding
if opt.mask:
out += tf.equal(opt.target, tf.zeros_like(opt.target)).sg_float()
return out
def ner_accuracy(tensor, opt):
r"""Returns accuracy of predictions.
Args:
tensor: A `Tensor`. Probability distributions or unscaled prediction scores.
opt:
target: A 'Tensor`. Labels.
Returns:
A `Tensor` of the same shape as `tensor`. Each value will be 1 if correct else 0.
For example,
```
tensor = [[20.1, 18, -4.2], [0.04, 21.1, 31.3]]
target = [[0, 1]]
tensor.sg_accuracy(target=target) => [[ 1. 0.]]
```
"""
assert opt.target is not None, 'target is mandatory.'
opt += tf.sg_opt(k=1)
# # calc accuracy
out = tf.identity(tf.equal(tensor.sg_argmax() + 1, tf.cast(opt.target, tf.int64)).sg_float(), name='acc')
# out = tf.identity(tf.nn.in_top_k(tensor, opt.target, opt.k).sg_float(), name='acc')
# masking padding
if opt.mask:
out += tf.equal(opt.target, tf.zeros_like(opt.target)).sg_float()
return out
def __init__(self, is_train=True):
# inputs
if is_train:
self.X, self.Y, self.num_batch = get_batch_data(
) # (16, 9, 9, 1), (16, 9, 9)
self.X_val, self.Y_val, _ = get_batch_data(is_train=False)
else:
self.X = tf.placeholder(tf.float32, [None, 9, 9, 1])
with tf.sg_context(size=3, act='relu', bn=True):
self.logits = self.X.sg_identity()
for _ in range(5):
self.logits = (self.logits.sg_conv(dim=512))
self.logits = self.logits.sg_conv(
dim=10, size=1, act='linear',
bn=False) # (16, 9, 9, 10) float32
if is_train:
self.ce = self.logits.sg_ce(target=self.Y,
mask=False) # (16, 9, 9) dtype=float32
self.istarget = tf.equal(
self.X.sg_squeeze(), tf.zeros_like(self.X.sg_squeeze())
).sg_float() # zeros: 1, non-zeros: 0 (16, 9, 9) dtype=float32
self.loss = self.ce * self.istarget # (16, 9, 9) dtype=float32
self.reduced_loss = self.loss.sg_sum() / self.istarget.sg_sum()
tf.sg_summary_loss(self.reduced_loss, "reduced_loss")
# accuracy evaluation ( for train set )
self.preds = (self.logits.sg_argmax()).sg_int()
self.hits = tf.equal(self.preds, self.Y).sg_float()
self.acc_train = (self.hits *
self.istarget).sg_sum() / self.istarget.sg_sum()
# accuracy evaluation ( for validation set )
self.preds_ = (self.logits.sg_reuse(
input=self.X_val).sg_argmax()).sg_int()
self.hits_ = tf.equal(self.preds_, self.Y_val).sg_float()
self.istarget_ = tf.equal(self.X_val.sg_squeeze(),
tf.zeros_like(
self.X_val.sg_squeeze())).sg_float()
self.acc_val = (self.hits_ *
self.istarget_).sg_sum() / self.istarget_.sg_sum()
def sg_accuracy(tensor, opt):
assert opt.target is not None, 'target is mandatory.'
opt += tf.sg_opt(k=1)
# # calc accuracy
out = tf.identity(tf.equal(tensor.sg_argmax(),
tf.cast(opt.target, tf.int64)).sg_float(),
name='acc')
# out = tf.identity(tf.nn.in_top_k(tensor, opt.target, opt.k).sg_float(), name='acc')
return out
def __init__(self, is_train=True):
# inputs
if is_train:
self.x, self.y, self.num_batch = get_batch_data()
self.x_val, self.y_val, _ = get_batch_data(is_train=False)
else:
self.x = tf.placeholder(tf.float32, [None, 9, 9, 1])
with tf.sg_context(size=3, act='relu', bn=True):
self.logits = self.x.sg_identity()
for _ in range(10):
self.logits = (self.logits.sg_conv(dim=512))
self.logits = self.logits.sg_conv(dim=10,
size=1,
act='linear',
bn=False)
if is_train:
self.ce = self.logits.sg_ce(target=self.y, mask=False)
self.istarget = tf.equal(self.x.sg_squeeze(),
tf.zeros_like(
self.x.sg_squeeze())).sg_float()
self.loss = self.ce * self.istarget
self.reduced_loss = self.loss.sg_sum() / self.istarget.sg_sum()
tf.sg_summary_loss(self.reduced_loss, "reduced_loss")
# accuracy evaluation ( for validation set )
self.preds_ = (self.logits.sg_reuse(
input=self.x_val).sg_argmax()).sg_int()
self.hits_ = tf.equal(self.preds_, self.y_val).sg_float()
self.istarget_ = tf.equal(self.x_val.sg_squeeze(),
tf.zeros_like(
self.x_val.sg_squeeze())).sg_float()
self.acc = (self.hits_ *
self.istarget_).sg_sum() / self.istarget_.sg_sum()
def testIt():
data = raw
positive = np.array(data.label_train) > 0
x = tf.placeholder(tf.float32, [None, 4096])
y = tf.placeholder(tf.float32)
disc_real = discriminator(x)
accuracy = tf.reduce_mean(
tf.cast(tf.equal(tf.cast(disc_real > 0.5, "float"), y), tf.float32))
np.set_printoptions(precision=3, suppress=True)
with tf.Session() as sess:
sess.run(
tf.group(tf.global_variables_initializer(),
tf.sg_phase().assign(False)))
# restore parameters
tf.sg_restore(sess,
tf.train.latest_checkpoint('asset/train/gan'),
category=['generator', 'discriminator'])
ans = sess.run(disc_real, feed_dict={x: np.array(data.test)})
print np.sum(ans > 0.5)
np.save('dm_bird.npy', ans)
def q_process(t1, t2):
'''
Processes each training sample so that it fits in the queue.
'''
# Lstrip zeros
zeros = tf.equal(t1, tf.zeros_like(t1)).sg_int().sg_sum()
t1 = t1[zeros:]
t2 = t2[zeros:]
# zero-PrePadding
t1 = tf.concat([tf.zeros([Hyperparams.seqlen-1], tf.int32), t1], 0)# 49 zero-prepadding
t2 = tf.concat([tf.zeros([Hyperparams.seqlen-1], tf.int32), t2], 0)# 49 zero-prepadding
# radom crop
stacked = tf.stack((t1, t2))
cropped = tf.random_crop(stacked, [2, Hyperparams.seqlen])
t1, t2 = cropped[0], cropped[1]
t2 = t2[-1]
return t1, t2
W = tf.Variable(
tf.random_normal([n_hidden_units_three, num_classes], mean=0, stddev=sd))
b = tf.Variable(tf.random_normal([num_classes], mean=0, stddev=sd))
with tf.name_scope('out'):
y_ = tf.nn.softmax(tf.matmul(h_3, W) + b, name="out")
init = tf.global_variables_initializer()
cost_function = tf.reduce_mean(
-tf.reduce_sum(Y * tf.log(y_), reduction_indices=[1]))
#optimizer = tf.train.RMSPropOptimizer(learning_rate,decay=0.9,momentum=0.9,centered=True).minimize(cost_function)
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(
cost_function)
correct_prediction = tf.equal(tf.argmax(y_, 1), tf.argmax(Y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
cost_history = np.empty(shape=[1], dtype=float)
acc_history = np.empty(shape=[1], dtype=float)
t_cost_history = np.empty(shape=[1], dtype=float)
t_acc_history = np.empty(shape=[1], dtype=float)
y_true, y_pred = None, None
with tf.Session() as session:
session.run(init)
saver = tf.train.Saver()
for epoch in range(epochs):
for batch in range(int(db_size / batchsize)):
indices = get_indices(batchsize)
feed = data_tools.next_minibatch(indices, db)