KyanChen's picture
init
f549064
# Copyright (c) OpenMMLab. All rights reserved.
import torch
import torch.nn as nn
from mmengine.model.weight_init import trunc_normal_
from mmcls.registry import MODELS
from .vision_transformer import VisionTransformer
@MODELS.register_module()
class DistilledVisionTransformer(VisionTransformer):
"""Distilled Vision Transformer.
A PyTorch implement of : `Training data-efficient image transformers &
distillation through attention <https://arxiv.org/abs/2012.12877>`_
Args:
arch (str | dict): Vision Transformer architecture. If use string,
choose from 'small', 'base', 'large', 'deit-tiny', 'deit-small'
and 'deit-base'. If use dict, it should have below keys:
- **embed_dims** (int): The dimensions of embedding.
- **num_layers** (int): The number of transformer encoder layers.
- **num_heads** (int): The number of heads in attention modules.
- **feedforward_channels** (int): The hidden dimensions in
feedforward modules.
Defaults to 'deit-base'.
img_size (int | tuple): The expected input image shape. Because we
support dynamic input shape, just set the argument to the most
common input image shape. Defaults to 224.
patch_size (int | tuple): The patch size in patch embedding.
Defaults to 16.
in_channels (int): The num of input channels. Defaults to 3.
out_indices (Sequence | int): Output from which stages.
Defaults to -1, means the last stage.
drop_rate (float): Probability of an element to be zeroed.
Defaults to 0.
drop_path_rate (float): stochastic depth rate. Defaults to 0.
qkv_bias (bool): Whether to add bias for qkv in attention modules.
Defaults to True.
norm_cfg (dict): Config dict for normalization layer.
Defaults to ``dict(type='LN')``.
final_norm (bool): Whether to add a additional layer to normalize
final feature map. Defaults to True.
with_cls_token (bool): Whether concatenating class token into image
tokens as transformer input. Defaults to True.
output_cls_token (bool): Whether output the cls_token. If set True,
``with_cls_token`` must be True. Defaults to True.
interpolate_mode (str): Select the interpolate mode for position
embeding vector resize. Defaults to "bicubic".
patch_cfg (dict): Configs of patch embeding. Defaults to an empty dict.
layer_cfgs (Sequence | dict): Configs of each transformer layer in
encoder. Defaults to an empty dict.
init_cfg (dict, optional): Initialization config dict.
Defaults to None.
"""
num_extra_tokens = 2 # cls_token, dist_token
def __init__(self, arch='deit-base', *args, **kwargs):
super(DistilledVisionTransformer, self).__init__(
arch=arch, *args, **kwargs)
self.dist_token = nn.Parameter(torch.zeros(1, 1, self.embed_dims))
def forward(self, x):
B = x.shape[0]
x, patch_resolution = self.patch_embed(x)
# stole cls_tokens impl from Phil Wang, thanks
cls_tokens = self.cls_token.expand(B, -1, -1)
dist_token = self.dist_token.expand(B, -1, -1)
x = torch.cat((cls_tokens, dist_token, x), dim=1)
x = x + self.resize_pos_embed(
self.pos_embed,
self.patch_resolution,
patch_resolution,
mode=self.interpolate_mode,
num_extra_tokens=self.num_extra_tokens)
x = self.drop_after_pos(x)
if not self.with_cls_token:
# Remove class token for transformer encoder input
x = x[:, 2:]
outs = []
for i, layer in enumerate(self.layers):
x = layer(x)
if i == len(self.layers) - 1 and self.final_norm:
x = self.norm1(x)
if i in self.out_indices:
B, _, C = x.shape
if self.with_cls_token:
patch_token = x[:, 2:].reshape(B, *patch_resolution, C)
patch_token = patch_token.permute(0, 3, 1, 2)
cls_token = x[:, 0]
dist_token = x[:, 1]
else:
patch_token = x.reshape(B, *patch_resolution, C)
patch_token = patch_token.permute(0, 3, 1, 2)
cls_token = None
dist_token = None
if self.output_cls_token:
out = [patch_token, cls_token, dist_token]
else:
out = patch_token
outs.append(out)
return tuple(outs)
def init_weights(self):
super(DistilledVisionTransformer, self).init_weights()
if not (isinstance(self.init_cfg, dict)
and self.init_cfg['type'] == 'Pretrained'):
trunc_normal_(self.dist_token, std=0.02)