From b379b5a5b3408cfe5641b39e6d4bf3307e09fcd0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E8=B0=A2=E6=98=95=E8=BE=B0?= <xinchen.xie@qq.com>
Date: Thu, 22 Apr 2021 11:19:55 +0800
Subject: [PATCH] support transformer backbone (#465)
* vit backbone
* fix lint
* add docstrings and fix pretrained pos_embed dim not match prob
* add unittest for vit
* fix lint
* add vit based fcn configs
* fix import error
* support multiple resolution input images
* upsample pos_embed at init_weights
* support resize pos_embed at evaluation
* fix training errors
* add more unitest code for vit backbone
* unitest for uncovered code
* add norm_eval unittest
* refactor _pos_embeding
* minor change
* change var name
* rafactor init_weight
* load weights after resize
* ignore 'module' in pretrain checkpoint
* add with_cp
* add with_cp
Co-authored-by: Jiarui XU <xvjiarui0826@gmail.com>
---
mmseg/models/backbones/__init__.py | 4 +-
mmseg/models/backbones/vit.py | 396 +++++++++++++++++++
tests/test_models/test_backbones/test_vit.py | 64 +++
3 files changed, 463 insertions(+), 1 deletion(-)
create mode 100644 mmseg/models/backbones/vit.py
create mode 100644 tests/test_models/test_backbones/test_vit.py
diff --git a/mmseg/models/backbones/__init__.py b/mmseg/models/backbones/__init__.py
index 740317da..eae064b6 100644
--- a/mmseg/models/backbones/__init__.py
+++ b/mmseg/models/backbones/__init__.py
@@ -7,8 +7,10 @@ from .resnest import ResNeSt
from .resnet import ResNet, ResNetV1c, ResNetV1d
from .resnext import ResNeXt
from .unet import UNet
+from .vit import VisionTransformer
__all__ = [
'ResNet', 'ResNetV1c', 'ResNetV1d', 'ResNeXt', 'HRNet', 'FastSCNN',
- 'ResNeSt', 'MobileNetV2', 'UNet', 'CGNet', 'MobileNetV3'
+ 'ResNeSt', 'MobileNetV2', 'UNet', 'CGNet', 'MobileNetV3',
+ 'VisionTransformer'
]
diff --git a/mmseg/models/backbones/vit.py b/mmseg/models/backbones/vit.py
new file mode 100644
index 00000000..bda2a354
--- /dev/null
+++ b/mmseg/models/backbones/vit.py
@@ -0,0 +1,396 @@
+"""Modified from https://github.com/rwightman/pytorch-image-
+models/blob/master/timm/models/vision_transformer.py."""
+
+import math
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.utils.checkpoint as cp
+from mmcv.cnn import (Conv2d, Linear, build_activation_layer, build_norm_layer,
+ constant_init, kaiming_init, normal_init, xavier_init)
+from mmcv.runner import _load_checkpoint
+from mmcv.utils.parrots_wrapper import _BatchNorm
+
+from mmseg.utils import get_root_logger
+from ..builder import BACKBONES
+
+
+class Mlp(nn.Module):
+ """MLP layer for Encoder block.
+
+ Args:
+ in_features(int): Input dimension for the first fully
+ connected layer.
+ hidden_features(int): Output dimension for the first fully
+ connected layer.
+ out_features(int): Output dementsion for the second fully
+ connected layer.
+ act_cfg(dict): Config dict for activation layer.
+ Default: dict(type='GELU').
+ drop(float): Drop rate for the dropout layer. Dropout rate has
+ to be between 0 and 1. Default: 0.
+ """
+
+ def __init__(self,
+ in_features,
+ hidden_features=None,
+ out_features=None,
+ act_cfg=dict(type='GELU'),
+ drop=0.):
+ super(Mlp, self).__init__()
+ out_features = out_features or in_features
+ hidden_features = hidden_features or in_features
+ self.fc1 = Linear(in_features, hidden_features)
+ self.act = build_activation_layer(act_cfg)
+ self.fc2 = Linear(hidden_features, out_features)
+ self.drop = nn.Dropout(drop)
+
+ def forward(self, x):
+ x = self.fc1(x)
+ x = self.act(x)
+ x = self.drop(x)
+ x = self.fc2(x)
+ x = self.drop(x)
+ return x
+
+
+class Attention(nn.Module):
+ """Attention layer for Encoder block.
+
+ Args:
+ dim (int): Dimension for the input vector.
+ num_heads (int): Number of parallel attention heads.
+ qkv_bias (bool): Enable bias for qkv if True. Default: False.
+ qk_scale (float): Override default qk scale of head_dim ** -0.5 if set.
+ attn_drop (float): Drop rate for attention output weights.
+ Default: 0.
+ proj_drop (float): Drop rate for output weights. Default: 0.
+ """
+
+ def __init__(self,
+ dim,
+ num_heads=8,
+ qkv_bias=False,
+ qk_scale=None,
+ attn_drop=0.,
+ proj_drop=0.):
+ super(Attention, self).__init__()
+ self.num_heads = num_heads
+ head_dim = dim // num_heads
+ self.scale = qk_scale or head_dim**-0.5
+
+ self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
+ self.attn_drop = nn.Dropout(attn_drop)
+ self.proj = Linear(dim, dim)
+ self.proj_drop = nn.Dropout(proj_drop)
+
+ def forward(self, x):
+ b, n, c = x.shape
+ qkv = self.qkv(x).reshape(b, n, 3, self.num_heads,
+ c // self.num_heads).permute(2, 0, 3, 1, 4)
+ q, k, v = qkv[0], qkv[1], qkv[2]
+
+ attn = (q @ k.transpose(-2, -1)) * self.scale
+ attn = attn.softmax(dim=-1)
+ attn = self.attn_drop(attn)
+
+ x = (attn @ v).transpose(1, 2).reshape(b, n, c)
+ x = self.proj(x)
+ x = self.proj_drop(x)
+ return x
+
+
+class Block(nn.Module):
+ """Implements encoder block with residual connection.
+
+ Args:
+ dim (int): The feature dimension.
+ num_heads (int): Number of parallel attention heads.
+ mlp_ratio (int): Ratio of mlp hidden dim to embedding dim.
+ qk_scale (float): Override default qk scale of head_dim ** -0.5 if set.
+ drop (float): Drop rate for mlp output weights. Default: 0.
+ attn_drop (float): Drop rate for attention output weights.
+ Default: 0.
+ proj_drop (float): Drop rate for attn layer output weights.
+ Default: 0.
+ act_cfg (dict): Config dict for activation layer.
+ Default: dict(type='GELU').
+ norm_cfg (dict): Config dict for normalization layer.
+ Default: dict(type='LN', requires_grad=True).
+ """
+
+ def __init__(self,
+ dim,
+ num_heads,
+ mlp_ratio=4,
+ qkv_bias=False,
+ qk_scale=None,
+ drop=0.,
+ attn_drop=0.,
+ proj_drop=0.,
+ act_cfg=dict(type='GELU'),
+ norm_cfg=dict(type='LN'),
+ with_cp=False):
+ super(Block, self).__init__()
+ self.with_cp = with_cp
+ _, self.norm1 = build_norm_layer(norm_cfg, dim)
+ self.attn = Attention(dim, num_heads, qkv_bias, qk_scale, attn_drop,
+ proj_drop)
+ _, self.norm2 = build_norm_layer(norm_cfg, dim)
+ mlp_hidden_dim = int(dim * mlp_ratio)
+ self.mlp = Mlp(
+ in_features=dim,
+ hidden_features=mlp_hidden_dim,
+ act_cfg=act_cfg,
+ drop=drop)
+
+ def forward(self, x):
+
+ def _inner_forward(x):
+ out = x + self.attn(self.norm1(x))
+ out = out + self.mlp(self.norm2(out))
+ return out
+
+ if self.with_cp and x.requires_grad:
+ out = cp.checkpoint(_inner_forward, x)
+ else:
+ out = _inner_forward(x)
+
+ return out
+
+
+class PatchEmbed(nn.Module):
+ """Image to Patch Embedding.
+
+ Args:
+ img_size (int， tuple): Input image size.
+ default: 224.
+ patch_size (int): Width and height for a patch.
+ default: 16.
+ in_channels (int): Input channels for images. Default: 3.
+ embed_dim (int): The embedding dimension. Default: 768.
+ """
+
+ def __init__(self,
+ img_size=224,
+ patch_size=16,
+ in_channels=3,
+ embed_dim=768):
+ super(PatchEmbed, self).__init__()
+ if isinstance(img_size, int):
+ self.img_size = (img_size, img_size)
+ elif isinstance(img_size, tuple):
+ self.img_size = img_size
+ else:
+ raise TypeError('img_size must be type of int or tuple')
+ h, w = self.img_size
+ self.patch_size = (patch_size, patch_size)
+ self.num_patches = (h // patch_size) * (w // patch_size)
+ self.proj = Conv2d(
+ in_channels, embed_dim, kernel_size=patch_size, stride=patch_size)
+
+ def forward(self, x):
+ return self.proj(x).flatten(2).transpose(1, 2)
+
+
+@BACKBONES.register_module()
+class VisionTransformer(nn.Module):
+ """Vision transformer backbone.
+
+ A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for
+ Image Recognition at Scale` - https://arxiv.org/abs/2010.11929
+
+ Args:
+ img_size (tuple): input image size. Default: (224, 224）.
+ patch_size (int, tuple): patch size. Default: 16.
+ in_channels (int): number of input channels. Default: 3.
+ embed_dim (int): embedding dimension. Default: 768.
+ depth (int): depth of transformer. Default: 12.
+ num_heads (int): number of attention heads. Default: 12.
+ mlp_ratio (int): ratio of mlp hidden dim to embedding dim. Default: 4.
+ qkv_bias (bool): enable bias for qkv if True. Default: True.
+ qk_scale (float): override default qk scale of head_dim ** -0.5 if set.
+ drop_rate (float): dropout rate. Default: 0.
+ attn_drop_rate (float): attention dropout rate. Default: 0.
+ norm_cfg (dict): Config dict for normalization layer.
+ Default: dict(type='LN', requires_grad=True).
+ act_cfg (dict): Config dict for activation layer.
+ Default: dict(type='GELU').
+ norm_eval (bool): Whether to set norm layers to eval mode, namely,
+ freeze running stats (mean and var). Note: Effect on Batch Norm
+ and its variants only. Default: False.
+ with_cp (bool): Use checkpoint or not. Using checkpoint
+ will save some memory while slowing down the training speed.
+ Default: False.
+ """
+
+ def __init__(self,
+ img_size=(224, 224),
+ patch_size=16,
+ in_channels=3,
+ embed_dim=768,
+ depth=12,
+ num_heads=12,
+ mlp_ratio=4,
+ qkv_bias=True,
+ qk_scale=None,
+ drop_rate=0.,
+ attn_drop_rate=0.,
+ norm_cfg=dict(type='LN'),
+ act_cfg=dict(type='GELU'),
+ norm_eval=False,
+ with_cp=False):
+ super(VisionTransformer, self).__init__()
+ self.img_size = img_size
+ self.patch_size = patch_size
+ self.features = self.embed_dim = embed_dim
+ self.patch_embed = PatchEmbed(
+ img_size=img_size,
+ patch_size=patch_size,
+ in_channels=in_channels,
+ embed_dim=embed_dim)
+
+ self.pos_embed = nn.Parameter(
+ torch.zeros(1, self.patch_embed.num_patches, embed_dim))
+ self.pos_drop = nn.Dropout(p=drop_rate)
+
+ self.blocks = nn.Sequential(*[
+ Block(
+ dim=embed_dim,
+ num_heads=num_heads,
+ mlp_ratio=mlp_ratio,
+ qkv_bias=qkv_bias,
+ qk_scale=qk_scale,
+ drop=drop_rate,
+ attn_drop=attn_drop_rate,
+ act_cfg=act_cfg,
+ norm_cfg=norm_cfg,
+ with_cp=with_cp) for i in range(depth)
+ ])
+ _, self.norm = build_norm_layer(norm_cfg, embed_dim)
+
+ self.norm_eval = norm_eval
+ self.with_cp = with_cp
+
+ def init_weights(self, pretrained=None):
+ if isinstance(pretrained, str):
+ logger = get_root_logger()
+ checkpoint = _load_checkpoint(pretrained, logger=logger)
+ if 'state_dict' in checkpoint:
+ state_dict = checkpoint['state_dict']
+ else:
+ state_dict = checkpoint
+
+ if 'pos_embed' in state_dict.keys():
+ state_dict['pos_embed'] = state_dict['pos_embed'][:, 1:, :]
+ logger.info(
+ msg='Remove the "cls_token" dimension from the checkpoint')
+
+ if self.pos_embed.shape != state_dict['pos_embed'].shape:
+ logger.info(msg=f'Resize the pos_embed shape from \
+ {state_dict["pos_embed"].shape} to \
+ {self.pos_embed.shape}')
+ h, w = self.img_size
+ pos_size = int(math.sqrt(state_dict['pos_embed'].shape[1]))
+ state_dict['pos_embed'] = self.resize_pos_embed(
+ state_dict['pos_embed'], (h, w), (pos_size, pos_size),
+ self.patch_size)
+ self.load_state_dict(state_dict, False)
+
+ elif pretrained is None:
+ # We only implement the 'jax_impl' initialization implemented at
+ # https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py#L353 # noqa: E501
+ normal_init(self.pos_embed)
+ for n, m in self.named_modules():
+ if isinstance(m, Linear):
+ xavier_init(m.weight, distribution='uniform')
+ if m.bias is not None:
+ if 'mlp' in n:
+ normal_init(m.bias, std=1e-6)
+ else:
+ constant_init(m.bias, 0)
+ elif isinstance(m, Conv2d):
+ kaiming_init(m.weight, mode='fan_in')
+ if m.bias is not None:
+ constant_init(m.bias, 0)
+ elif isinstance(m, (_BatchNorm, nn.GroupNorm, nn.LayerNorm)):
+ constant_init(m.bias, 0)
+ constant_init(m.weight, 1)
+ else:
+ raise TypeError('pretrained must be a str or None')
+
+ def _pos_embeding(self, img, patched_img, pos_embed):
+ """Positiong embeding method.
+
+ Resize the pos_embed, if the input image size doesn't match
+ the training size.
+ Args:
+ img (torch.Tensor): The inference image tensor, the shape
+ must be [B, C, H, W].
+ patched_img (torch.Tensor): The patched image, it should be
+ shape of [B, L1, C].
+ pos_embed (torch.Tensor): The pos_embed weighs, it should be
+ shape of [B, L2, c].
+ Return:
+ torch.Tensor: The pos encoded image feature.
+ """
+ assert patched_img.ndim == 3 and pos_embed.ndim == 3, \
+ 'the shapes of patched_img and pos_embed must be [B, L, C]'
+ x_len, pos_len = patched_img.shape[1], pos_embed.shape[1]
+ if x_len != pos_len:
+ if pos_len == (self.img_size[0] // self.patch_size) * (
+ self.img_size[1] // self.patch_size):
+ pos_h = self.img_size[0] // self.patch_size
+ pos_w = self.img_size[1] // self.patch_size
+ else:
+ raise ValueError(
+ 'Unexpected shape of pos_embed, got {}.'.format(
+ pos_embed.shape))
+ pos_embed = self.resize_pos_embed(pos_embed, img.shape[2:],
+ (pos_h, pos_w), self.patch_size)
+ return patched_img + pos_embed
+
+ @staticmethod
+ def resize_pos_embed(pos_embed, input_shpae, pos_shape, patch_size):
+ """Resize pos_embed weights.
+
+ Resize pos_embed using bicubic interpolate method.
+ Args:
+ pos_embed (torch.Tensor): pos_embed weights.
+ input_shpae (tuple): Tuple for (input_h, intput_w).
+ pos_shape (tuple): Tuple for (pos_h, pos_w).
+ patch_size (int): Patch size.
+ Return:
+ torch.Tensor: The resized pos_embed of shape [B, L_new, C]
+ """
+ assert pos_embed.ndim == 3, 'shape of pos_embed must be [B, L, C]'
+ input_h, input_w = input_shpae
+ pos_h, pos_w = pos_shape
+ pos_embed = pos_embed.reshape(1, pos_h, pos_w,
+ pos_embed.shape[2]).permute(0, 3, 1, 2)
+ pos_embed = F.interpolate(
+ pos_embed,
+ size=[input_h // patch_size, input_w // patch_size],
+ align_corners=False,
+ mode='bicubic')
+ pos_embed = torch.flatten(pos_embed, 2).transpose(1, 2)
+ return pos_embed
+
+ def forward(self, inputs):
+ x = self.patch_embed(inputs)
+ x = self._pos_embeding(inputs, x, self.pos_embed)
+ x = self.blocks(x)
+ x = self.norm(x)
+ B, _, C = x.shape
+ x = x.reshape(B, inputs.shape[2] // self.patch_size,
+ inputs.shape[3] // self.patch_size,
+ C).permute(0, 3, 1, 2)
+ return [x]
+
+ def train(self, mode=True):
+ super(VisionTransformer, self).train(mode)
+ if mode and self.norm_eval:
+ for m in self.modules():
+ if isinstance(m, nn.LayerNorm):
+ m.eval()
diff --git a/tests/test_models/test_backbones/test_vit.py b/tests/test_models/test_backbones/test_vit.py
new file mode 100644
index 00000000..5c5572e4
--- /dev/null
+++ b/tests/test_models/test_backbones/test_vit.py
@@ -0,0 +1,64 @@
+import pytest
+import torch
+
+from mmseg.models.backbones.vit import VisionTransformer
+from .utils import check_norm_state
+
+
+def test_vit_backbone():
+ with pytest.raises(TypeError):
+ # pretrained must be a string path
+ model = VisionTransformer()
+ model.init_weights(pretrained=0)
+
+ with pytest.raises(TypeError):
+ # img_size must be int or tuple
+ model = VisionTransformer(img_size=512.0)
+
+ with pytest.raises(TypeError):
+ # test upsample_pos_embed function
+ x = torch.randn(1, 196)
+ VisionTransformer.resize_pos_embed(x, 512, 512, 224, 224)
+
+ with pytest.raises(RuntimeError):
+ # forward inputs must be [N, C, H, W]
+ x = torch.randn(3, 30, 30)
+ model = VisionTransformer()
+ model(x)
+
+ # Test img_size isinstance int
+ imgs = torch.randn(1, 3, 224, 224)
+ model = VisionTransformer(img_size=224)
+ model.init_weights()
+ model(imgs)
+
+ # Test norm_eval = True
+ model = VisionTransformer(norm_eval=True)
+ model.train()
+
+ # Test ViT backbone with input size of 224 and patch size of 16
+ model = VisionTransformer()
+ model.init_weights()
+ model.train()
+
+ assert check_norm_state(model.modules(), True)
+
+ # Test large size input image
+ imgs = torch.randn(1, 3, 256, 256)
+ feat = model(imgs)
+ assert feat[0].shape == (1, 768, 16, 16)
+
+ # Test small size input image
+ imgs = torch.randn(1, 3, 32, 32)
+ feat = model(imgs)
+ assert feat[0].shape == (1, 768, 2, 2)
+
+ imgs = torch.randn(1, 3, 224, 224)
+ feat = model(imgs)
+ assert feat[0].shape == (1, 768, 14, 14)
+
+ # Test with_cp=True
+ model = VisionTransformer(with_cp=True)
+ imgs = torch.randn(1, 3, 224, 224)
+ feat = model(imgs)
+ assert feat[0].shape == (1, 768, 14, 14)
--
GitLab