def __init__(self, in_channels, out_channels, kernel_size, stride=1,
pad=0, nobias=False):
"""畳込みレイヤ
Parameters
----------
in_channels : int or None
入力データのチャンネル数。Noneの場合はforward時のxからin_channelsを取得する
out_channels : int
出力データのチャンネル数
kernel_size : int or (int, int)
：カーネルサイズ
stride : int or (int, int)
ストライド
pad : int or (int, int)
パディング
nobias : bool
バイアスを使用するかどうか
"""
super().__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.pad = pad
self.W = Parameter(None, name='W')
if nobias:
self.b = None
else:
b_data = np.zeros(out_channels).astype(np.float32)
self.b = Parameter(b_data, name='b')
def __init__(self, in_channels, out_channels, kernel_size, stride=1,
pad=0, nobias=False, dtype=np.float32):
"""Two-dimensional convolutional layer.
Args:
in_channels (int or None): Number of channels of input arrays. If
`None`, parameter initialization will be deferred until the first
forward data pass at which time the size will be determined.
out_channels (int): Number of channels of output arrays.
kernel_size (int or (int, int)): Size of filters.
stride (int or (int, int)): Stride of filter applications.
pad (int or (int, int)): Spatial padding width for input arrays.
nobias (bool): If `True`, then this function does not use the bias.
"""
super().__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.pad = pad
self.dtype = dtype
self.W = Parameter(None, name='W')
if in_channels is not None:
self._init_W()
if nobias:
self.b = None
else:
self.b = Parameter(np.zeros(out_channels, dtype=dtype), name='b')
def __init__(self, in_size, out_size, nobias=False, dtype=np.float32):
super().__init__()
I, O = in_size, out_size
W_data = np.random.randn(I, O).astype(dtype) * np.sqrt(1 / I)
self.W = Parameter(W_data, name='W')
if nobias:
self.b = None
else:
self.b = Parameter(np.zeros(O, dtype=dtype), name='b')
def __init__(self):
super().__init__()
# `.avg_mean` and `.avg_var` are `Parameter` objects, so they will be
# saved to a file (using `save_weights()`).
# But they don't need grads, so they're just used as `ndarray`.
self.avg_mean = Parameter(None, name='avg_mean')
self.avg_var = Parameter(None, name='avg_var')
self.gamma = Parameter(None, name='gamma')
self.beta = Parameter(None, name='beta')
def __init__(self, out_size, nobias=False, dtype=np.float32, in_size=None):
super().__init__()
self.in_size = in_size
self.out_size = out_size
self.dtype = dtype
self.W = Parameter(None, name='W')
if self.in_size is not None:
self.__init_W()
if nobias:
self.b = None
else:
self.b = Parameter(np.zeros(out_size, dtype=dtype), name='b')
def __init__(self, in_size, out_size=None, nobias=False):
super().__init__()
if out_size is None:
in_size, out_size = None, in_size
self.in_size = in_size
self.out_size = out_size
self.W = Parameter(None, name='W')
if nobias:
self.b = None
else:
self.b = Parameter(np.zeros(out_size, dtype=np.float32), name='b')
def __init__(self,
out_size: int,
nobias: bool = False,
dtype=np.float32,
in_size: Optional[int] = None) -> None:
super().__init__()
self.in_size = in_size
self.out_size = out_size
self.dtype = dtype
self.W = Parameter(None, name="W")
if self.in_size is not None:
self._init_W()
if nobias:
self.b = None
else:
self.b = Parameter(np.zeros(out_size, dtype=dtype), name='b')
def __init__(self,
in_channels,
out_channels,
kernel_size,
stride=1,
pad=0,
nobias=False):
super().__init__()
self.kernel_size = _pair(kernel_size)
self.stride = stride
self.pad = pad
I, O = in_channels, out_channels
KH, KW = self.kernel_size
W_data = np.random.randn(O, I, KH, KW).astype('f') * np.sqrt(
1 / I * KH * KW)
self.W = Parameter(W_data, name='W')
if nobias:
self.b = None
else:
self.b = Parameter(np.zeros(O).astype('f'), name='b')
def __init__(self,
out_channels,
kernel_size,
stride=1,
pad=0,
nobias=False,
dtype=np.float32,
in_channels=None):
super().__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.pad = pad
self.dtype = dtype
self.W = Parameter(None, name='W')
if in_channels is not None:
self._init_W()
if nobias:
self.b = None
else:
self.b = Parameter(np.zeros(out_channels, dtype=dtype), name='b')
def __init__(self, in_size, out_size):
super().__init__()
self.W = Parameter(np.random.randn(in_size, out_size), name='W')
y = F.sigmoid(y)
y = l2(y)
return y
lr = 0.2
iters = 10000
for i in range(iters):
y_pred = predict(x)
loss = F.mean_squared_error(y, y_pred)
l1.cleargrads()
l2.cleargrads()
loss.backward()
for l in [l1, l2]:
for p in l.params():
p.data -= lr * p.grad.data
if i % 1000 == 0:
print(loss)
# ここからはテスト
from dezero.core import Parameter
layer = L.Layer()
layer.p1 = Parameter(np.array(1))
print(type(layer.p1))
print(layer._params)
print(layer.p1)
for name in layer._params:
print(name, layer.__dict__[name])
import numpy as np from dezero.core import Parameter from dezero.layers import TestClass test = TestClass() test.p1 = Parameter(np.array(1.0)) test.p2 = Parameter(np.array(2.0))
