#!/usr/bin/env python3
import pdb
import gym
import torch as th
import torch.nn as nn
import torchvision
from stable_baselines3.common.torch_layers import BaseFeaturesExtractor
[docs]class CustomResnet10CNN(BaseFeaturesExtractor):
"""
:param observation_space: (gym.Space)
:param features_dim: (int) Number of features extracted.
This corresponds to the number of unit for the last layer.
"""
def __init__(self, observation_space: gym.spaces.Box, features_dim: int = 256):
super(CustomResnet10CNN, self).__init__(observation_space, features_dim)
# We assume CxHxW images (channels first)
# Re-ordering will be done by pre-preprocessing or wrapper
n_input_channels = observation_space.shape[0]
self.cnn = _resnet(BasicBlock, [2, 2, 2, 2],num_channels = n_input_channels)
with th.no_grad():
n_flatten = self.cnn(
th.as_tensor(observation_space.sample()[None]).float()
).shape[1]
self.linear = nn.Sequential(nn.Linear(n_flatten, features_dim), nn.ReLU())
[docs] def forward(self, observations: th.Tensor) -> th.Tensor:
# Cut off image
# reshape to from vector to W*H
# gray to color transform
# application of ResNet
# Concat features to the rest of observation vector
# return
return self.linear(self.cnn(observations))
def conv3x3(in_planes, out_planes, stride=1, groups=1, dilation=1):
"""3x3 convolution with padding"""
return nn.Conv2d(
in_planes,
out_planes,
kernel_size=3,
stride=stride,
padding=dilation,
groups=groups,
bias=False,
dilation=dilation
)
def conv1x1(in_planes, out_planes, stride=1):
"""1x1 convolution"""
return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)
class BasicBlock(nn.Module):
expansion = 1
def __init__(
self, inplanes, planes, stride=1, downsample=None, groups=1, base_width=64, dilation=1, norm_layer=None
):
super(BasicBlock, self).__init__()
if norm_layer is None:
norm_layer = nn.BatchNorm2d
if groups != 1 or base_width != 64:
raise ValueError('BasicBlock only supports groups=1 and base_width=64')
if dilation > 1:
raise NotImplementedError("Dilation > 1 not supported in BasicBlock")
# Both self.conv1 and self.downsample layers downsample the input when stride != 1
# layers inside each basic block of a residual block
self.conv1 = conv3x3(inplanes, planes, stride)
self.bn1 = norm_layer(planes)
self.relu = nn.ReLU(inplace=True)
self.conv2 = conv3x3(planes, planes)
self.bn2 = norm_layer(planes)
# last two are operations and not layers
self.downsample = downsample
self.stride = stride
def forward(self, x):
# saving x to pass over the bridge connection
identity = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
if self.downsample is not None:
identity = self.downsample(x)
out += identity
out = self.relu(out)
return out
class ResNet(nn.Module):
def __init__(
self,
block,
layers,
num_channels,
num_classes=1000,
zero_init_residual=False,
groups=1,
width_per_group=64,
replace_stride_with_dilation=None,
norm_layer=None,
return_all_feature_maps=False,
first_conv=True, # pre-processing layers which makes the image size half [64->32]
maxpool1=True # used in pre-processing
):
super(ResNet, self).__init__()
if norm_layer is None:
norm_layer = nn.BatchNorm2d
self._norm_layer = norm_layer
self.return_all_feature_maps = return_all_feature_maps
self.inplanes = 64
self.dilation = 1
if replace_stride_with_dilation is None:
# each element in the tuple indicates if we should replace
# the 2x2 stride with a dilated convolution instead
replace_stride_with_dilation = [False, False, False]
if len(replace_stride_with_dilation) != 3:
raise ValueError(
"replace_stride_with_dilation should be None "
f"or a 3-element tuple, got {replace_stride_with_dilation}"
)
self.groups = groups
self.base_width = width_per_group
# ------ layers before first residual block ---------------
if first_conv:
self.conv1 = nn.Conv2d(num_channels, self.inplanes, kernel_size=7, stride=2, padding=3, bias=False)
else:
self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=3, stride=1, padding=1, bias=False)
self.bn1 = norm_layer(self.inplanes)
self.relu = nn.ReLU(inplace=True)
if maxpool1:
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
else:
self.maxpool = nn.MaxPool2d(kernel_size=1, stride=1)
# ------ residual blocks start here ------------------------
# BLOCK - 1
self.layer1 = self._make_layer(block, 64, layers[0])
# BLOCK - 2
self.layer2 = self._make_layer(block, 512, layers[1], stride=2, dilate=replace_stride_with_dilation[0])
self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
self.fc = nn.Linear(512 * block.expansion, num_classes)
for m in self.modules():
if isinstance(m, nn.Conv2d):
nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
nn.init.constant_(m.weight, 1)
nn.init.constant_(m.bias, 0)
# Zero-initialize the last BN in each residual branch,
# so that the residual branch starts with zeros, and each residual block behaves like an identity.
# This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677
if zero_init_residual:
for m in self.modules():
if isinstance(m, Bottleneck):
nn.init.constant_(m.bn3.weight, 0)
elif isinstance(m, BasicBlock):
nn.init.constant_(m.bn2.weight, 0)
def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
norm_layer = self._norm_layer
downsample = None
previous_dilation = self.dilation
if dilate:
self.dilation *= stride
stride = 1
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
conv1x1(self.inplanes, planes * block.expansion, stride),
norm_layer(planes * block.expansion),
)
layers = []
layers.append(
block(
self.inplanes,
planes,
stride,
downsample,
self.groups,
self.base_width,
previous_dilation,
norm_layer,
)
)
self.inplanes = planes * block.expansion
for _ in range(1, blocks):
layers.append(
block(
self.inplanes,
planes,
groups=self.groups,
base_width=self.base_width,
dilation=self.dilation,
norm_layer=norm_layer
)
)
return nn.Sequential(*layers)
def forward(self, x):
# passing input from pre-processing layers
x0 = self.conv1(x)
x0 = self.bn1(x0)
x0 = self.relu(x0)
x0 = self.maxpool(x0)
# passing input from residual blocks
if self.return_all_feature_maps:
x1 = self.layer1(x0) # block1
x2 = self.layer2(x1) # block2
return [x0, x1, x2]
else:
x0 = self.layer1(x0)
x0 = self.layer2(x0)
x0 = self.avgpool(x0)
x0 = th.flatten(x0, 1)
return x0
def _resnet(block, layers, **kwargs):
model = ResNet(block, layers, **kwargs)
model.fc = nn.Identity()
return model