Function Differences with tf.random.uniform_candidate_sampler

tf.random.uniform_candidate_sampler

tf.random.uniform_candidate_sampler(
    true_classes,
    num_true,
    num_sampled,
    unique,
    range_max,
    seed=None,
    name=None
)(sampled_candidates, true_expected_count, sampled_expected_count)  -> Tuple

For more information, see tf.random.uniform_candidate_sampler.

mindspore.ops.uniform_candidate_sampler

mindspore.ops.uniform_candidate_sampler(
    true_classes,
    num_true,
    num_sampled,
    unique,
    range_max,
    seed=0,
    remove_accidental_hits=False
)(sampled_candidates, true_expected_count, sampled_expected_count) -> Tuple

For more information, see mindspore.ops.uniform_candidate_sampler.

Differences

TensorFlow: Sample a set of internal aliases using a uniform distribution and return three Tensors.

MindSpore: MindSpore API implements the same functions as TensorFlow, with some parameter names different.

Categories	Subcategories	TensorFlow	MindSpore	Differences
Parameters	Parameter 1	true_classes	true_classes	-
	Parameter 2	num_true	num_true	-
	Parameter 3	num_sampled	num_sampled	-
	Parameter 4	unique	unique	-
	Parameter 5	range_max	range_max	-
	Parameter 6	seed	seed	-
	Parameter 7	-	remove_accidental_hits	Indicates whether to remove the accidental hit. Default: False
	Parameter 8	name	-	Not involved
Return Parameters	Parameter 1	sampled_candidates	sampled_candidates	-
	Parameter 2	true_expected_count	true_expected_count	-
	Parameter 3	sampled_expected_count	sampled_expected_count	-

Code Example 1

The outputs of MindSpore and TensorFlow are consistent.

# TensorFlow
import tensorflow as tf
import numpy as np

data = tf.constant(np.random.rand(5, 3), dtype=tf.int64)
out1, out2, out3 = tf.random.uniform_candidate_sampler(data, 3, 3, False, 5, 0)
print(out1.shape)
# (3,)
print(out2.shape)
# (5, 3)
print(out3.shape)
# (3,)

# MindSpore
import mindspore
import numpy as np
from mindspore.ops import function as ops
from mindspore import Tensor

data = Tensor(np.random.rand(5, 3), mindspore.int64)
out1, out2, out3 = ops.uniform_candidate_sampler(data, 3, 3, False, 5, 0)
print(out1.shape)
# (3,)
print(out2.shape)
# (5, 3)
print(out3.shape)
# (3,)