funasr/models/encoder/opennmt_encoders/conv_encoder.py
@@ -67,7 +67,7 @@ class ConvEncoder(AbsEncoder): """ author: Speech Lab, Alibaba Group, China Author: Speech Lab of DAMO Academy, Alibaba Group Convolution encoder in OpenNMT framework """ @@ -137,12 +137,12 @@ self.out_padding = nn.ConstantPad1d((left_padding, right_padding), 0.0) self.conv_out = nn.Conv1d( num_units, num_units, out_units, kernel_size, ) if self.out_norm: self.after_norm = LayerNorm(num_units) self.after_norm = LayerNorm(out_units) def output_size(self) -> int: return self.num_units
