blob: d56955e52dc3f60975aee1d08c7f1b6c53900b79 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
|
"""Conformer module."""
from copy import deepcopy
from torch import nn, Tensor
from text_recognizer.networks.conformer.block import ConformerBlock
class Conformer(nn.Module):
def __init__(self, block: ConformerBlock, depth: int) -> None:
super().__init__()
self.blocks = nn.ModuleList([deepcopy(block) for _ in range(depth)])
def forward(self, x: Tensor) -> Tensor:
for fn in self.blocks:
x = fn(x)
return x
|