Function Differences with tf.compat.v1.layers.Dense
tf.compat.v1.layers.Dense
class tf.compat.v1.layers.Dense(
units,
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=tf.compat.v1.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs
)(x) -> Tensor
For more information, see tf.compat.v1.layers.Dense.
mindspore.nn.Dense
class mindspore.nn.Dense(
in_channels,
out_channels,
weight_init='normal',
bias_init='zeros',
has_bias=True,
activation=None
)(x) -> Tensor
For more information, see mindspore.nn.Dense.
Differences
TensorFlow: Fully connected layer that implements the matrix multiplication operation.
MindSpore: MindSpore API basically implements the same function as TensorFlow.
Categories |
Subcategories |
TensorFlow |
MindSpore |
Differences |
---|---|---|---|---|
Parameters |
Parameter 1 |
units |
out_channels |
Same function, different parameter names |
Parameter 2 |
activation |
activation |
- |
|
Parameter 3 |
use_bias |
has_bias |
Same function, different parameter names |
|
Parameter 4 |
kernel_initializer |
weight_init |
Same function, different parameter names |
|
Parameter 5 |
bias_initializer |
bias_init |
Same function, different parameter names |
|
Parameter 6 |
kernel_regularizer |
- |
The regular function of the weight matrix. MindSpore does not have this parameter. |
|
Parameter 7 |
bias_regularizer |
- |
The regularization function for the deviation. MindSpore does not have this parameter. |
|
Parameter 8 |
activity_regularizer |
- |
The regularization function for the output. MindSpore does not have this parameter. |
|
Parameter 9 |
kernel_constraint |
- |
Optional projection functions that will be applied to the kernel after the |
|
Parameter 10 |
bias_constraint |
- |
Optional projection function to be applied to the deviation after being updated by |
|
Parameter 11 |
trainable |
- |
Boolean. If |
|
Parameter 12 |
name |
- |
Not involved |
|
Parameter 13 |
**kwargs |
- |
Not involved |
|
Parameter 14 |
- |
in_channels |
The spatial dimension of the input. TensorFlow does not have this parameter |
|
Input |
Single input |
x |
x |
- |
Code Example
The two APIs achieve the same function and have the same usage.
# TensorFlow
import tensorflow as tf
from tensorflow.compat.v1 import layers
import numpy as np
model = layers.Dense(4)
x = tf.constant(np.array([[180, 234, 154], [244, 48, 247]]),dtype=tf.float32)
output = model(x)
print(output.shape)
# (2, 4)
# MindSpore
import mindspore
from mindspore import Tensor, nn
import numpy as np
x = Tensor(np.array([[180, 234, 154], [244, 48, 247]]), mindspore.float32)
net = nn.Dense(3, 4)
output = net(x)
print(output.shape)
# (2, 4)