File size: 2,925 Bytes
a89d9fd |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 |
# copyright (c) 2019 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import math
import paddle
from paddle import ParamAttr, nn
from paddle.nn import functional as F
def get_para_bias_attr(l2_decay, k):
regularizer = paddle.regularizer.L2Decay(l2_decay)
stdv = 1.0 / math.sqrt(k * 1.0)
initializer = nn.initializer.Uniform(-stdv, stdv)
weight_attr = ParamAttr(regularizer=regularizer, initializer=initializer)
bias_attr = ParamAttr(regularizer=regularizer, initializer=initializer)
return [weight_attr, bias_attr]
class CTCHead(nn.Layer):
def __init__(self,
in_channels,
out_channels,
fc_decay=0.0004,
mid_channels=None,
return_feats=False,
**kwargs):
super(CTCHead, self).__init__()
if mid_channels is None:
weight_attr, bias_attr = get_para_bias_attr(
l2_decay=fc_decay, k=in_channels)
self.fc = nn.Linear(
in_channels,
out_channels,
weight_attr=weight_attr,
bias_attr=bias_attr)
else:
weight_attr1, bias_attr1 = get_para_bias_attr(
l2_decay=fc_decay, k=in_channels)
self.fc1 = nn.Linear(
in_channels,
mid_channels,
weight_attr=weight_attr1,
bias_attr=bias_attr1)
weight_attr2, bias_attr2 = get_para_bias_attr(
l2_decay=fc_decay, k=mid_channels)
self.fc2 = nn.Linear(
mid_channels,
out_channels,
weight_attr=weight_attr2,
bias_attr=bias_attr2)
self.out_channels = out_channels
self.mid_channels = mid_channels
self.return_feats = return_feats
def forward(self, x, targets=None):
if self.mid_channels is None:
predicts = self.fc(x)
else:
x = self.fc1(x)
predicts = self.fc2(x)
if self.return_feats:
result = (x, predicts)
else:
result = predicts
if not self.training:
predicts = F.softmax(predicts, axis=2)
result = predicts
return result
|