80 lines
2.8 KiB
Python
80 lines
2.8 KiB
Python
""" Padding Helpers
|
|
|
|
Hacked together by / Copyright 2020 Ross Wightman
|
|
"""
|
|
import math
|
|
from typing import List, Tuple
|
|
|
|
import torch
|
|
import torch.nn.functional as F
|
|
|
|
|
|
# Calculate symmetric padding for a convolution
|
|
def get_padding(kernel_size: int, stride: int = 1, dilation: int = 1, **_) -> int:
|
|
padding = ((stride - 1) + dilation * (kernel_size - 1)) // 2
|
|
return padding
|
|
|
|
|
|
# Calculate asymmetric TensorFlow-like 'SAME' padding for a convolution
|
|
def get_same_padding(x: int, kernel_size: int, stride: int, dilation: int):
|
|
if isinstance(x, torch.Tensor):
|
|
return torch.clamp(((x / stride).ceil() - 1) * stride + (kernel_size - 1) * dilation + 1 - x, min=0)
|
|
else:
|
|
return max((math.ceil(x / stride) - 1) * stride + (kernel_size - 1) * dilation + 1 - x, 0)
|
|
|
|
|
|
# Can SAME padding for given args be done statically?
|
|
def is_static_pad(kernel_size: int, stride: int = 1, dilation: int = 1, **_):
|
|
return stride == 1 and (dilation * (kernel_size - 1)) % 2 == 0
|
|
|
|
|
|
def pad_same_arg(
|
|
input_size: List[int],
|
|
kernel_size: List[int],
|
|
stride: List[int],
|
|
dilation: List[int] = (1, 1),
|
|
) -> List[int]:
|
|
ih, iw = input_size
|
|
kh, kw = kernel_size
|
|
pad_h = get_same_padding(ih, kh, stride[0], dilation[0])
|
|
pad_w = get_same_padding(iw, kw, stride[1], dilation[1])
|
|
return [pad_w // 2, pad_w - pad_w // 2, pad_h // 2, pad_h - pad_h // 2]
|
|
|
|
|
|
# Dynamically pad input x with 'SAME' padding for conv with specified args
|
|
def pad_same(
|
|
x,
|
|
kernel_size: List[int],
|
|
stride: List[int],
|
|
dilation: List[int] = (1, 1),
|
|
value: float = 0,
|
|
):
|
|
ih, iw = x.size()[-2:]
|
|
pad_h = get_same_padding(ih, kernel_size[0], stride[0], dilation[0])
|
|
pad_w = get_same_padding(iw, kernel_size[1], stride[1], dilation[1])
|
|
x = F.pad(x, (pad_w // 2, pad_w - pad_w // 2, pad_h // 2, pad_h - pad_h // 2), value=value)
|
|
return x
|
|
|
|
|
|
def get_padding_value(padding, kernel_size, **kwargs) -> Tuple[Tuple, bool]:
|
|
dynamic = False
|
|
if isinstance(padding, str):
|
|
# for any string padding, the padding will be calculated for you, one of three ways
|
|
padding = padding.lower()
|
|
if padding == 'same':
|
|
# TF compatible 'SAME' padding, has a performance and GPU memory allocation impact
|
|
if is_static_pad(kernel_size, **kwargs):
|
|
# static case, no extra overhead
|
|
padding = get_padding(kernel_size, **kwargs)
|
|
else:
|
|
# dynamic 'SAME' padding, has runtime/GPU memory overhead
|
|
padding = 0
|
|
dynamic = True
|
|
elif padding == 'valid':
|
|
# 'VALID' padding, same as padding=0
|
|
padding = 0
|
|
else:
|
|
# Default to PyTorch style 'same'-ish symmetric padding
|
|
padding = get_padding(kernel_size, **kwargs)
|
|
return padding, dynamic
|