| /external/tensorflow/tensorflow/contrib/opt/python/training/ |
| sign_decay.py | 28 from tensorflow.python.ops import math_ops
53 global_step = math_ops.minimum(global_step, decay_steps)
54 remaining_steps = math_ops.to_int32(decay_steps) - math_ops.to_int32(
56 decayed = math_ops.to_float(remaining_steps) / math_ops.to_float(
58 return math_ops.maximum(0.0, decayed)
94 global_step = math_ops.minimum(global_step, decay_steps)
95 completed_fraction = math_ops.to_float(global_step) / math_ops.to_float
[all...] |
| lazy_adam_optimizer.py | 30 from tensorflow.python.ops import math_ops
51 beta1_power = math_ops.cast(beta1_power, var.dtype.base_dtype)
52 beta2_power = math_ops.cast(beta2_power, var.dtype.base_dtype)
53 lr_t = math_ops.cast(self._lr_t, var.dtype.base_dtype)
54 beta1_t = math_ops.cast(self._beta1_t, var.dtype.base_dtype)
55 beta2_t = math_ops.cast(self._beta2_t, var.dtype.base_dtype)
56 epsilon_t = math_ops.cast(self._epsilon_t, var.dtype.base_dtype)
57 lr = (lr_t * math_ops.sqrt(1 - beta2_power) / (1 - beta1_power))
70 (1 - beta2_t) * math_ops.square(grad.values),
76 denominator_slice = math_ops.sqrt(v_t_slice) + epsilon_
[all...] |
| nadam_optimizer.py | 22 from tensorflow.python.ops import math_ops
42 math_ops.cast(beta1_power, var.dtype.base_dtype),
43 math_ops.cast(beta2_power, var.dtype.base_dtype),
44 math_ops.cast(self._lr_t, var.dtype.base_dtype),
45 math_ops.cast(self._beta1_t, var.dtype.base_dtype),
46 math_ops.cast(self._beta2_t, var.dtype.base_dtype),
47 math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
60 math_ops.cast(beta1_power, grad.dtype.base_dtype),
61 math_ops.cast(beta2_power, grad.dtype.base_dtype),
62 math_ops.cast(self._lr_t, grad.dtype.base_dtype)
[all...] |
| /external/tensorflow/tensorflow/python/training/ |
| learning_rate_decay.py | 26 from tensorflow.python.ops import math_ops
98 global_step = math_ops.cast(global_step, dtype)
99 decay_steps = math_ops.cast(decay_steps, dtype)
100 decay_rate = math_ops.cast(decay_rate, dtype)
103 p = math_ops.floor(p)
104 return math_ops.multiply(
105 learning_rate, math_ops.pow(decay_rate, p), name=name)
161 b = math_ops.cast(b, x.dtype.base_dtype)
275 global_step = math_ops.cast(global_step, dtype)
276 decay_steps = math_ops.cast(decay_steps, dtype
[all...] |
| ftrl.py | 22 from tensorflow.python.ops import math_ops
152 math_ops.cast(self._learning_rate_tensor, var.dtype.base_dtype),
153 math_ops.cast(self._l1_regularization_strength_tensor,
155 math_ops.cast(self._l2_regularization_strength_tensor,
157 math_ops.cast(self._learning_rate_power_tensor, var.dtype.base_dtype),
165 math_ops.cast(self._learning_rate_tensor, var.dtype.base_dtype),
166 math_ops.cast(self._l1_regularization_strength_tensor,
168 math_ops.cast(self._l2_regularization_strength_tensor,
170 math_ops.cast(self._l2_shrinkage_regularization_strength_tensor,
172 math_ops.cast(self._learning_rate_power_tensor, var.dtype.base_dtype)
[all...] |
| rmsprop.py | 47 from tensorflow.python.ops import math_ops
140 math_ops.cast(self._learning_rate_tensor, var.dtype.base_dtype),
141 math_ops.cast(self._decay_tensor, var.dtype.base_dtype),
142 math_ops.cast(self._momentum_tensor, var.dtype.base_dtype),
143 math_ops.cast(self._epsilon_tensor, var.dtype.base_dtype),
151 math_ops.cast(self._learning_rate_tensor, var.dtype.base_dtype),
152 math_ops.cast(self._decay_tensor, var.dtype.base_dtype),
153 math_ops.cast(self._momentum_tensor, var.dtype.base_dtype),
154 math_ops.cast(self._epsilon_tensor, var.dtype.base_dtype),
168 math_ops.cast(self._learning_rate_tensor, grad.dtype.base_dtype)
[all...] |
| adadelta.py | 22 from tensorflow.python.ops import math_ops
77 math_ops.cast(self._lr_t, var.dtype.base_dtype),
78 math_ops.cast(self._rho_t, var.dtype.base_dtype),
79 math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
90 math_ops.cast(self._lr_t, grad.dtype.base_dtype),
91 math_ops.cast(self._rho_t, grad.dtype.base_dtype),
92 math_ops.cast(self._epsilon_t, grad.dtype.base_dtype),
103 math_ops.cast(self._lr_t, var.dtype.base_dtype),
104 math_ops.cast(self._rho_t, var.dtype.base_dtype),
105 math_ops.cast(self._epsilon_t, var.dtype.base_dtype)
[all...] |
| /external/tensorflow/tensorflow/contrib/metrics/python/metrics/ |
| classification.py | 24 from tensorflow.python.ops import math_ops
57 is_correct = math_ops.cast(
58 math_ops.equal(predictions, labels), dtypes.float32)
60 is_correct = math_ops.multiply(is_correct, weights)
61 num_values = math_ops.multiply(weights, array_ops.ones_like(is_correct))
62 return math_ops.div(math_ops.reduce_sum(is_correct),
63 math_ops.reduce_sum(num_values))
64 return math_ops.reduce_mean(is_correct)
|
| /external/tensorflow/tensorflow/contrib/losses/python/losses/ |
| loss_ops.py | 27 from tensorflow.python.ops import math_ops
62 reduced_losses = math_ops.reduce_sum(
64 reduced_losses = math_ops.multiply(reduced_losses, weights)
65 return math_ops.reduce_sum(reduced_losses)
85 math_ops.greater(denominator, 0),
86 math_ops.div(numerator,
88 math_ops.equal(denominator, 0),
105 total_loss = math_ops.reduce_sum(losses)
129 losses = math_ops.to_float(losses)
130 weights = math_ops.to_float(ops.convert_to_tensor(weights)
[all...] |
| /external/tensorflow/tensorflow/contrib/sparsemax/python/ops/ |
| sparsemax.py | 24 from tensorflow.python.ops import math_ops
52 z = logits - math_ops.reduce_mean(logits, axis=1)[:, array_ops.newaxis]
58 z_cumsum = math_ops.cumsum(z_sorted, axis=1)
59 k = math_ops.range(
60 1, math_ops.cast(dims, logits.dtype) + 1, dtype=logits.dtype)
64 k_z = math_ops.reduce_sum(math_ops.cast(z_check, dtypes.int32), axis=1)
67 indices = array_ops.stack([math_ops.range(0, obs), k_z - 1], axis=1)
69 tau_z = (tau_sum - 1) / math_ops.cast(k_z, logits.dtype)
72 return math_ops.maximum
[all...] |
| /external/tensorflow/tensorflow/contrib/layers/python/ops/ |
| bucketization_op.py | 20 from tensorflow.python.ops import math_ops
40 return math_ops._bucketize( # pylint: disable=protected-access
|
| /external/tensorflow/tensorflow/contrib/losses/python/metric_learning/ |
| metric_loss_ops.py | 27 from tensorflow.python.ops import math_ops
52 pairwise_distances_squared = math_ops.add(
53 math_ops.reduce_sum(
54 math_ops.square(feature),
57 math_ops.reduce_sum(
58 math_ops.square(
61 keepdims=True)) - 2.0 * math_ops.matmul(
65 pairwise_distances_squared = math_ops.maximum(pairwise_distances_squared, 0.0)
67 error_mask = math_ops.less_equal(pairwise_distances_squared, 0.0)
73 pairwise_distances = math_ops.sqrt
[all...] |
| /external/tensorflow/tensorflow/python/kernel_tests/ |
| bincount_op_test.py | 15 """Tests for math_ops.bincount."""
25 from tensorflow.python.ops import math_ops
34 math_ops.bincount([], minlength=5).eval(), [0, 0, 0, 0, 0])
35 self.assertAllEqual(math_ops.bincount([], minlength=1).eval(), [0])
36 self.assertAllEqual(math_ops.bincount([], minlength=0).eval(), [])
38 math_ops.bincount([], minlength=0, dtype=np.float32).eval().dtype,
41 math_ops.bincount([], minlength=3, dtype=np.float64).eval().dtype,
47 math_ops.bincount([1, 1, 1, 2, 2, 3]).eval(), [0, 3, 2, 1])
49 self.assertAllEqual(math_ops.bincount(arr).eval(), [0, 5, 4, 3, 2, 1])
51 self.assertAllEqual(math_ops.bincount(arr).eval(), [6, 5, 4, 3, 2, 1]
[all...] |
| basic_gpu_test.py | 33 from tensorflow.python.ops import math_ops
60 self._compareGPU(x, y, np.add, math_ops.add)
61 self._compareGPU(x, y, np.subtract, math_ops.subtract)
62 self._compareGPU(x, y, np.multiply, math_ops.multiply)
63 self._compareGPU(x, y + 0.1, np.true_divide, math_ops.truediv)
64 self._compareGPU(x, y + 0.1, np.floor_divide, math_ops.floordiv)
65 self._compareGPU(x, y, np.power, math_ops.pow)
70 self._compareGPU(x, y, np.add, math_ops.add)
71 self._compareGPU(x, y, np.subtract, math_ops.subtract)
72 self._compareGPU(x, y, np.multiply, math_ops.multiply
[all...] |
| /external/tensorflow/tensorflow/python/ops/ |
| spectral_grad.py | 25 from tensorflow.python.ops import math_ops
30 return math_ops.reduce_prod(array_ops.shape(grad)[-rank:])
35 size = math_ops.cast(_FFTSizeForGrad(grad, 1), dtypes.float32)
36 return spectral_ops.ifft(grad) * math_ops.complex(size, 0.)
41 rsize = 1. / math_ops.cast(_FFTSizeForGrad(grad, 1), dtypes.float32)
42 return spectral_ops.fft(grad) * math_ops.complex(rsize, 0.)
47 size = math_ops.cast(_FFTSizeForGrad(grad, 2), dtypes.float32)
48 return spectral_ops.ifft2d(grad) * math_ops.complex(size, 0.)
53 rsize = 1. / math_ops.cast(_FFTSizeForGrad(grad, 2), dtypes.float32)
54 return spectral_ops.fft2d(grad) * math_ops.complex(rsize, 0.
[all...] |
| math_grad.py | 15 """Gradients for operators defined in math_ops.py."""
30 from tensorflow.python.ops import math_ops
35 return x // math_ops.maximum(y, 1)
61 output_shape_kept_dims = math_ops.reduced_shape(input_shape, op.inputs[1])
70 output_shape_kept_dims = math_ops.reduced_shape(input_shape, op.inputs[1])
78 indicators = math_ops.cast(math_ops.equal(y, op.inputs[0]), grad.dtype)
80 math_ops.reduce_sum(indicators, op.inputs[1]), output_shape_kept_dims)
82 return [math_ops.div(indicators, num_selected) * grad, None]
112 math_ops.reduce_prod(input_shape), math_ops.reduce_prod(output_shape)
[all...] |
| /external/tensorflow/tensorflow/contrib/distributions/python/ops/bijectors/ |
| sinh_arcsinh.py | 27 from tensorflow.python.ops import math_ops
38 math_ops.abs(x) * np.sqrt(np.finfo(x.dtype.as_numpy_dtype).eps) <= 1.,
39 math_ops.sqrt(x**2. + 1.),
55 math_ops.abs(x))
141 return math_ops.sinh((math_ops.asinh(x) + self.skewness) * self.tailweight)
144 return math_ops.sinh(math_ops.asinh(y) / self.tailweight - self.skewness)
153 return math_ops.reduce_sum(
156 math_ops.log(math_ops.cosh
[all...] |
| sigmoid.py | 21 from tensorflow.python.ops import math_ops
39 return math_ops.sigmoid(x)
42 return math_ops.log(y) - math_ops.log1p(-y)
45 return -math_ops.log(y) - math_ops.log1p(-y)
|
| /external/tensorflow/tensorflow/contrib/distributions/python/ops/ |
| normal_conjugate_posteriors.py | 21 from tensorflow.python.ops import math_ops
75 n = math_ops.cast(n, prior.dtype)
76 scale0_2 = math_ops.square(prior.scale)
77 scale_2 = math_ops.square(scale)
81 scale=math_ops.sqrt(scalep_2))
141 n = math_ops.cast(n, prior.dtype)
142 scale0_2 = math_ops.square(prior.scale)
143 scale_2 = math_ops.square(scale)
147 scale=math_ops.sqrt(scalep_2 + scale_2))
|
| /external/tensorflow/tensorflow/contrib/signal/python/ops/ |
| window_ops.py | 29 from tensorflow.python.ops import math_ops
112 periodic = math_ops.cast(
116 even = 1 - math_ops.mod(window_length, 2)
118 n = math_ops.cast(window_length + periodic * even - 1, dtype=dtype)
119 count = math_ops.cast(math_ops.range(window_length), dtype)
123 return math_ops.cast(a - b * math_ops.cos(cos_arg), dtype=dtype)
125 math_ops.equal(window_length, 1),
127 lambda: math_ops.cast(a - b * math_ops.cos(cos_arg), dtype=dtype)
[all...] |
| /external/tensorflow/tensorflow/python/ops/losses/ |
| losses_impl.py | 25 from tensorflow.python.ops import math_ops
92 math_ops.greater(denominator, 0),
93 math_ops.div(numerator, array_ops.where(
94 math_ops.equal(denominator, 0),
111 total_loss = math_ops.reduce_sum(losses)
138 weights = math_ops.to_float(weights)
140 math_ops.equal(weights, 0.0),
145 return math_ops.reduce_sum(
146 present, axis=math_ops.range(1, array_ops.rank(present)),
148 return math_ops.reduce_sum(present, name=scope
[all...] |
| /external/tensorflow/tensorflow/contrib/timeseries/python/timeseries/state_space_models/ |
| periodic.py | 30 from tensorflow.python.ops import math_ops
76 math_ops.range(self._periodicity - 1, dtype=powers.dtype),
83 is_row_negative = math_ops.equal(range_shape_padded + 1, powers[..., None])
92 is_one = math_ops.equal(coord_diff % self._periodicity,
101 return math_ops.cast(positive_ones + negative_row_indicator[..., None],
141 math_ops.range(self._periodicity, dtype=num_steps.dtype),
153 self.dtype)[..., None] * noise_addition_scalar * math_ops.cast(
265 value = math_ops.cast(value, self.dtype)
266 return math_ops.less(
267 math_ops.abs(value - gen_math_ops.round(value))
[all...] |
| /external/tensorflow/tensorflow/contrib/linear_optimizer/python/ops/ |
| sdca_ops.py | 32 from tensorflow.python.ops import math_ops
212 math_ops.reduce_sum(
213 math_ops.abs(math_ops.cast(weights, dtypes.float64))))
215 return self._options['symmetric_l1_regularization'] * math_ops.add_n(sums)
225 math_ops.reduce_sum(
226 math_ops.square(math_ops.cast(weights, dtypes.float64))))
228 return l2 * math_ops.add_n(sums) / 2.0
242 result_sparse += math_ops.segment_sum
[all...] |
| /external/tensorflow/tensorflow/contrib/boosted_trees/python/utils/ |
| losses.py | 24 from tensorflow.python.ops import math_ops
40 labels = math_ops.to_float(labels)
68 labels = math_ops.to_int64(labels)
75 labels = math_ops.reduce_sum(
77 labels = math_ops.to_float(labels)
80 unnormalized_probs = math_ops.exp(logits)
81 normalizers = math_ops.reduce_sum(unnormalized_probs, 1, keepdims=True)
82 softmax_predictions = math_ops.divide(unnormalized_probs,
83 math_ops.add(normalizers, eps))
86 probs_for_real_class = math_ops.reduce_sum(labels * softmax_predictions, 1
[all...] |
| /external/tensorflow/tensorflow/contrib/bayesflow/python/ops/ |
| csiszar_divergence_impl.py | 47 from tensorflow.python.ops import math_ops
108 f = math_ops.exp(logu) * logu
110 f = math_ops.expm1(alpha * logu) / (alpha * (alpha - 1.))
116 return f + math_ops.expm1(logu)
118 return f - math_ops.expm1(logu)
120 return f - math_ops.expm1(logu) / (alpha - 1.)
274 return math_ops.exp(logu) * logu - (1. + math_ops.exp(logu)) * y
327 return (1. + math_ops.exp(logu)) * y
359 return 0.5 * math_ops.abs(math_ops.expm1(logu)
[all...] |
