core/model/backbone/vit_inflora.py

""" Vision Transformer (ViT) in PyTorch
A PyTorch implement of Vision Transformers as described in:
'An Image Is Worth 16 x 16 Words: Transformers for Image Recognition at Scale'
    - https://arxiv.org/abs/2010.11929
`How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers`
    - https://arxiv.org/abs/2106.10270
The official jax code is released and available at https://github.com/google-research/vision_transformer
Acknowledgments:
* The paper authors for releasing code and weights, thanks!
* I fixed my class token impl based on Phil Wang's https://github.com/lucidrains/vit-pytorch ... check it out
for some einops/einsum fun
* Simple transformer style inspired by Andrej Karpathy's https://github.com/karpathy/minGPT
* Bert reference code checks against Huggingface Transformers and Tensorflow Bert
Hacked together by / Copyright 2020, Ross Wightman
"""

import math
import logging
from functools import partial
from collections import OrderedDict

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint

from timm.data import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD, IMAGENET_INCEPTION_MEAN, IMAGENET_INCEPTION_STD
from timm.models.helpers import build_model_with_cfg, resolve_pretrained_cfg, named_apply, adapt_input_conv, checkpoint_seq
from timm.models.layers import PatchEmbed, Mlp, DropPath, trunc_normal_, lecun_normal_
from timm.models.registry import register_model

_logger = logging.getLogger(__name__)


def _cfg(url='', **kwargs):
    return {
        'url': url,
        'num_classes': 1000, 'input_size': (3, 224, 224), 'pool_size': None,
        'crop_pct': .9, 'interpolation': 'bicubic', 'fixed_input_size': True,
        'mean': IMAGENET_INCEPTION_MEAN, 'std': IMAGENET_INCEPTION_STD,
        'first_conv': 'patch_embed.proj', 'classifier': 'head',
        **kwargs
    }


default_cfgs = {
    # patch models (weights from official Google JAX impl)
    'vit_tiny_patch16_224': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'Ti_16-i21k-300ep-lr_0.001-aug_none-wd_0.03-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.03-res_224.npz'),
    'vit_tiny_patch16_384': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'Ti_16-i21k-300ep-lr_0.001-aug_none-wd_0.03-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.03-res_384.npz',
        input_size=(3, 384, 384), crop_pct=1.0),
    'vit_small_patch32_224': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'S_32-i21k-300ep-lr_0.001-aug_light1-wd_0.03-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.03-res_224.npz'),
    'vit_small_patch32_384': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'S_32-i21k-300ep-lr_0.001-aug_light1-wd_0.03-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.03-res_384.npz',
        input_size=(3, 384, 384), crop_pct=1.0),
    'vit_small_patch16_224': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'S_16-i21k-300ep-lr_0.001-aug_light1-wd_0.03-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.03-res_224.npz'),
    'vit_small_patch16_384': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'S_16-i21k-300ep-lr_0.001-aug_light1-wd_0.03-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.03-res_384.npz',
        input_size=(3, 384, 384), crop_pct=1.0),
    'vit_base_patch32_224': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'B_32-i21k-300ep-lr_0.001-aug_medium1-wd_0.03-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.03-res_224.npz'),
    'vit_base_patch32_384': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'B_32-i21k-300ep-lr_0.001-aug_light1-wd_0.1-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.03-res_384.npz',
        input_size=(3, 384, 384), crop_pct=1.0),
    'vit_base_patch16_224': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'B_16-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.01-res_224.npz'),
    'vit_base_patch16_384': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'B_16-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.01-res_384.npz',
        input_size=(3, 384, 384), crop_pct=1.0),
    'vit_base_patch8_224': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'B_8-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.01-res_224.npz'),
    'vit_large_patch32_224': _cfg(
        url='',  # no official model weights for this combo, only for in21k
        ),
    'vit_large_patch32_384': _cfg(
        url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_p32_384-9b920ba8.pth',
        input_size=(3, 384, 384), crop_pct=1.0),
    'vit_large_patch16_224': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'L_16-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.1-sd_0.1--imagenet2012-steps_20k-lr_0.01-res_224.npz'),
    'vit_large_patch16_384': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/'
            'L_16-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.1-sd_0.1--imagenet2012-steps_20k-lr_0.01-res_384.npz',
        input_size=(3, 384, 384), crop_pct=1.0),

    'vit_large_patch14_224': _cfg(url=''),
    'vit_huge_patch14_224': _cfg(url=''),
    'vit_giant_patch14_224': _cfg(url=''),
    'vit_gigantic_patch14_224': _cfg(url=''),

    'vit_base2_patch32_256': _cfg(url='', input_size=(3, 256, 256), crop_pct=0.95),

    # patch models, imagenet21k (weights from official Google JAX impl)
    'vit_tiny_patch16_224_in21k': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/Ti_16-i21k-300ep-lr_0.001-aug_none-wd_0.03-do_0.0-sd_0.0.npz',
        num_classes=21843),
    'vit_small_patch32_224_in21k': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/S_32-i21k-300ep-lr_0.001-aug_light1-wd_0.03-do_0.0-sd_0.0.npz',
        num_classes=21843),
    'vit_small_patch16_224_in21k': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/S_16-i21k-300ep-lr_0.001-aug_light1-wd_0.03-do_0.0-sd_0.0.npz',
        num_classes=21843),
    'vit_base_patch32_224_in21k': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/B_32-i21k-300ep-lr_0.001-aug_medium1-wd_0.03-do_0.0-sd_0.0.npz',
        num_classes=21843),
    'vit_base_patch16_224_in21k': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/B_16-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.0-sd_0.0.npz',
        num_classes=21843),
    'vit_base_patch8_224_in21k': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/B_8-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.0-sd_0.0.npz',
        num_classes=21843),
    'vit_large_patch32_224_in21k': _cfg(
        url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_patch32_224_in21k-9046d2e7.pth',
        num_classes=21843),
    'vit_large_patch16_224_in21k': _cfg(
        url='https://storage.googleapis.com/vit_models/augreg/L_16-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.1-sd_0.1.npz',
        num_classes=21843),
    'vit_huge_patch14_224_in21k': _cfg(
        url='https://storage.googleapis.com/vit_models/imagenet21k/ViT-H_14.npz',
        hf_hub_id='timm/vit_huge_patch14_224_in21k',
        num_classes=21843),

    # SAM trained models (https://arxiv.org/abs/2106.01548)
    'vit_base_patch32_224_sam': _cfg(
        url='https://storage.googleapis.com/vit_models/sam/ViT-B_32.npz'),
    'vit_base_patch16_224_sam': _cfg(
        url='https://storage.googleapis.com/vit_models/sam/ViT-B_16.npz'),

    # DINO pretrained - https://arxiv.org/abs/2104.14294 (no classifier head, for fine-tune only)
    'vit_small_patch16_224_dino': _cfg(
        url='https://dl.fbaipublicfiles.com/dino/dino_deitsmall16_pretrain/dino_deitsmall16_pretrain.pth',
        mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD, num_classes=0),
    'vit_small_patch8_224_dino': _cfg(
        url='https://dl.fbaipublicfiles.com/dino/dino_deitsmall8_pretrain/dino_deitsmall8_pretrain.pth',
        mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD, num_classes=0),
    'vit_base_patch16_224_dino': _cfg(
        url='https://dl.fbaipublicfiles.com/dino/dino_vitbase16_pretrain/dino_vitbase16_pretrain.pth',
        mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD, num_classes=0),
    'vit_base_patch8_224_dino': _cfg(
        url='https://dl.fbaipublicfiles.com/dino/dino_vitbase8_pretrain/dino_vitbase8_pretrain.pth',
        mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD, num_classes=0),


    # ViT ImageNet-21K-P pretraining by MILL
    'vit_base_patch16_224_miil_in21k': _cfg(
        url='https://miil-public-eu.oss-eu-central-1.aliyuncs.com/model-zoo/ImageNet_21K_P/models/timm/vit_base_patch16_224_in21k_miil.pth',
        mean=(0, 0, 0), std=(1, 1, 1), crop_pct=0.875, interpolation='bilinear', num_classes=11221,
    ),
    'vit_base_patch16_224_miil': _cfg(
        url='https://miil-public-eu.oss-eu-central-1.aliyuncs.com/model-zoo/ImageNet_21K_P/models/timm'
            '/vit_base_patch16_224_1k_miil_84_4.pth',
        mean=(0, 0, 0), std=(1, 1, 1), crop_pct=0.875, interpolation='bilinear',
    ),

    # experimental
    'vit_small_patch16_36x1_224': _cfg(url=''),
    'vit_small_patch16_18x2_224': _cfg(url=''),
    'vit_base_patch16_18x2_224': _cfg(url=''),
}


class Attention_LoRA(nn.Module):
    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0., r=64, n_tasks=10):
        super().__init__()

        self.num_heads = num_heads
        head_dim = dim // num_heads
        # NOTE scale factor was wrong in my original version, can set manually to be compat with prev weights
        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 = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)
        self.attn_gradients = None
        self.attention_map = None
        self.rank = r

        self.lora_A_k = nn.ModuleList([nn.Linear(dim, r, bias=False) for _ in range(n_tasks)])
        self.lora_B_k = nn.ModuleList([nn.Linear(r, dim, bias=False) for _ in range(n_tasks)])
        self.lora_A_v = nn.ModuleList([nn.Linear(dim, r, bias=False) for _ in range(n_tasks)])
        self.lora_B_v = nn.ModuleList([nn.Linear(r, dim, bias=False) for _ in range(n_tasks)])
        self.rank = r

        self.matrix = torch.zeros(dim ,dim)
        self.n_matrix = 0
        self.cur_matrix = torch.zeros(dim ,dim)
        self.n_cur_matrix = 0

    def init_param(self):
        for t in range(len(self.lora_A_k)):
            nn.init.kaiming_uniform_(self.lora_A_k[t].weight, a=math.sqrt(5))
            nn.init.kaiming_uniform_(self.lora_A_v[t].weight, a=math.sqrt(5))
            nn.init.zeros_(self.lora_B_k[t].weight)
            nn.init.zeros_(self.lora_B_v[t].weight)

    def save_attn_gradients(self, attn_gradients):
        self.attn_gradients = attn_gradients
        
    def get_attn_gradients(self):
        return self.attn_gradients
    
    def save_attention_map(self, attention_map):
        self.attention_map = attention_map
        
    def get_attention_map(self):
        return self.attention_map
    
    def forward(self, x, task, register_hook=False, get_feat=False,get_cur_feat=False):
        if get_feat:
            self.matrix = (self.matrix*self.n_matrix + torch.bmm(x.detach().permute(0, 2, 1), x.detach()).sum(dim=0).cpu())/(self.n_matrix + x.shape[0]*x.shape[1])
            self.n_matrix += x.shape[0]*x.shape[1]
        if get_cur_feat:
            self.cur_matrix = (self.cur_matrix*self.n_cur_matrix + torch.bmm(x.detach().permute(0, 2, 1), x.detach()).sum(dim=0).cpu())/(self.n_cur_matrix + x.shape[0]*x.shape[1])
            self.n_cur_matrix += x.shape[0]*x.shape[1]

        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]   # make torchscript happy (cannot use tensor as tuple)

        # insert lora
        if task > -0.5:
            weight_k = torch.stack([torch.mm(self.lora_B_k[t].weight, self.lora_A_k[t].weight) for t in range(task+1)], dim=0).sum(dim=0)
            weight_v = torch.stack([torch.mm(self.lora_B_v[t].weight, self.lora_A_v[t].weight) for t in range(task+1)], dim=0).sum(dim=0)
            k = k + F.linear(x, weight_k).reshape(B, N, self.num_heads, C // self.num_heads).permute(0, 2, 1, 3)
            v = v + F.linear(x, weight_v).reshape(B, N, self.num_heads, C // self.num_heads).permute(0, 2, 1, 3)

        attn = (q @ k.transpose(-2, -1)) * self.scale
        attn = attn.softmax(dim=-1)
        attn = self.attn_drop(attn)
                
        if register_hook:
            self.save_attention_map(attn)
            attn.register_hook(self.save_attn_gradients)        

        x = (attn @ v).transpose(1, 2).reshape(B, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x

    def get_matrix(self, task):
        matrix_k = torch.mm(self.lora_B_k[task].weight, self.lora_A_k[task].weight)
        matrix_v = torch.mm(self.lora_B_v[task].weight, self.lora_A_v[task].weight)
        return matrix_k, matrix_v
    
    def get_pre_matrix(self, task):
        with torch.no_grad():
            weight_k = torch.stack([torch.mm(self.lora_B_k[t].weight, self.lora_A_k[t].weight) for t in range(task)], dim=0).sum(dim=0)
            weight_v = torch.stack([torch.mm(self.lora_B_v[t].weight, self.lora_A_v[t].weight) for t in range(task)], dim=0).sum(dim=0)
        return weight_k, weight_v

class LayerScale(nn.Module):
    def __init__(self, dim, init_values=1e-5, inplace=False):
        super().__init__()
        self.inplace = inplace
        self.gamma = nn.Parameter(init_values * torch.ones(dim))

    def forward(self, x):
        return x.mul_(self.gamma) if self.inplace else x * self.gamma


class Block(nn.Module):

    def __init__(
            self, dim, num_heads, mlp_ratio=4., qkv_bias=False, drop=0., attn_drop=0., init_values=None,
            drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm, n_tasks=10, r=64):
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention_LoRA(dim, num_heads=num_heads, qkv_bias=qkv_bias, attn_drop=attn_drop, proj_drop=drop, n_tasks=n_tasks, r=r)
        self.ls1 = LayerScale(dim, init_values=init_values) if init_values else nn.Identity()
        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
        self.drop_path1 = DropPath(drop_path) if drop_path > 0. else nn.Identity()

        self.norm2 = norm_layer(dim)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)
        self.ls2 = LayerScale(dim, init_values=init_values) if init_values else nn.Identity()
        self.drop_path2 = DropPath(drop_path) if drop_path > 0. else nn.Identity()

    def forward(self, x, task, register_hook=False, get_feat=False, get_cur_feat=False):
        x = x + self.drop_path1(self.ls1(self.attn(self.norm1(x), task, register_hook=register_hook, get_feat=get_feat, get_cur_feat=get_cur_feat)))
        x = x + self.drop_path2(self.ls2(self.mlp(self.norm2(x))))
        return x


class ParallelBlock(nn.Module):

    def __init__(
            self, dim, num_heads, num_parallel=2, mlp_ratio=4., qkv_bias=False, init_values=None,
            drop=0., attn_drop=0., drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm):
        super().__init__()
        self.num_parallel = num_parallel
        self.attns = nn.ModuleList()
        self.ffns = nn.ModuleList()
        for _ in range(num_parallel):
            self.attns.append(nn.Sequential(OrderedDict([
                ('norm', norm_layer(dim)),
                ('attn', Attention_LoRA(dim, num_heads=num_heads, qkv_bias=qkv_bias, attn_drop=attn_drop, proj_drop=drop)),
                ('ls', LayerScale(dim, init_values=init_values) if init_values else nn.Identity()),
                ('drop_path', DropPath(drop_path) if drop_path > 0. else nn.Identity())
            ])))
            self.ffns.append(nn.Sequential(OrderedDict([
                ('norm', norm_layer(dim)),
                ('mlp', Mlp(dim, hidden_features=int(dim * mlp_ratio), act_layer=act_layer, drop=drop)),
                ('ls', LayerScale(dim, init_values=init_values) if init_values else nn.Identity()),
                ('drop_path', DropPath(drop_path) if drop_path > 0. else nn.Identity())
            ])))

    def _forward_jit(self, x):
        x = x + torch.stack([attn(x) for attn in self.attns]).sum(dim=0)
        x = x + torch.stack([ffn(x) for ffn in self.ffns]).sum(dim=0)
        return x

    @torch.jit.ignore
    def _forward(self, x):
        x = x + sum(attn(x) for attn in self.attns)
        x = x + sum(ffn(x) for ffn in self.ffns)
        return x

    def forward(self, x):
        if torch.jit.is_scripting() or torch.jit.is_tracing():
            return self._forward_jit(x)
        else:
            return self._forward(x)


class VisionTransformer(nn.Module):
    """ Vision Transformer
    A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale`
        - https://arxiv.org/abs/2010.11929
    """

    def __init__(
            self, img_size=224, patch_size=16, in_chans=3, num_classes=1000, global_pool='token',
            embed_dim=768, depth=12, num_heads=12, mlp_ratio=4., qkv_bias=True, representation_size=None,
            drop_rate=0., attn_drop_rate=0., drop_path_rate=0., weight_init='', init_values=None,
            embed_layer=PatchEmbed, norm_layer=None, act_layer=None, block_fn=Block, n_tasks=10, rank=64):
        """
        Args:
            img_size (int, tuple): input image size
            patch_size (int, tuple): patch size
            in_chans (int): number of input channels
            num_classes (int): number of classes for classification head
            global_pool (str): type of global pooling for final sequence (default: 'token')
            embed_dim (int): embedding dimension
            depth (int): depth of transformer
            num_heads (int): number of attention heads
            mlp_ratio (int): ratio of mlp hidden dim to embedding dim
            qkv_bias (bool): enable bias for qkv if True
            representation_size (Optional[int]): enable and set representation layer (pre-logits) to this value if set
            drop_rate (float): dropout rate
            attn_drop_rate (float): attention dropout rate
            drop_path_rate (float): stochastic depth rate
            weight_init: (str): weight init scheme
            init_values: (float): layer-scale init values
            embed_layer (nn.Module): patch embedding layer
            norm_layer: (nn.Module): normalization layer
            act_layer: (nn.Module): MLP activation layer
        """
        super().__init__()
        assert global_pool in ('', 'avg', 'token')
        norm_layer = norm_layer or partial(nn.LayerNorm, eps=1e-6)
        act_layer = act_layer or nn.GELU

        self.num_classes = num_classes
        self.global_pool = global_pool
        self.num_features = self.embed_dim = embed_dim  # num_features for consistency with other models
        self.num_tokens = 1
        self.grad_checkpointing = False

        self.patch_embed = embed_layer(
            img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim)
        num_patches = self.patch_embed.num_patches

        self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
        self.cls_token_grow = nn.Parameter(torch.zeros(1, 5000, embed_dim))
        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + self.num_tokens, embed_dim))
        self.pos_embed_grow = nn.Parameter(torch.zeros(1, num_patches + 1000, embed_dim))
        self.pos_drop = nn.Dropout(p=drop_rate)

        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]  # stochastic depth decay rule
        self.blocks = nn.Sequential(*[
            block_fn(
                dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, init_values=init_values,
                drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[i], norm_layer=norm_layer, act_layer=act_layer,n_tasks=n_tasks,r=rank)
            for i in range(depth)])
        use_fc_norm = self.global_pool == 'avg'
        self.norm = norm_layer(embed_dim) if not use_fc_norm else nn.Identity()

        # Representation layer. Used for original ViT models w/ in21k pretraining.
        self.representation_size = representation_size
        self.pre_logits = nn.Identity()
        if representation_size:
            self._reset_representation(representation_size)

        # Classifier Head
        self.fc_norm = norm_layer(embed_dim) if use_fc_norm else nn.Identity()
        final_chs = self.representation_size if self.representation_size else self.embed_dim
        self.head = nn.Linear(final_chs, num_classes) if num_classes > 0 else nn.Identity()
        self.out_dim = final_chs

        if weight_init != 'skip':
            self.init_weights(weight_init)

    def _reset_representation(self, representation_size):
        self.representation_size = representation_size
        if self.representation_size:
            self.pre_logits = nn.Sequential(OrderedDict([
                ('fc', nn.Linear(self.embed_dim, self.representation_size)),
                ('act', nn.Tanh())
            ]))
        else:
            self.pre_logits = nn.Identity()

    def init_weights(self, mode=''):
        assert mode in ('jax', 'jax_nlhb', 'moco', '')
        head_bias = -math.log(self.num_classes) if 'nlhb' in mode else 0.
        trunc_normal_(self.pos_embed, std=.02)
        trunc_normal_(self.pos_embed_grow, std=.02)
        nn.init.normal_(self.cls_token, std=1e-6)
        nn.init.normal_(self.cls_token_grow, std=1e-6)
        named_apply(get_init_weights_vit(mode, head_bias), self)

    def _init_weights(self, m):
        # this fn left here for compat with downstream users
        init_weights_vit_timm(m)

    @torch.jit.ignore()
    def load_pretrained(self, checkpoint_path, prefix=''):
        _load_weights(self, checkpoint_path, prefix)

    @torch.jit.ignore
    def no_weight_decay(self):
        return {'pos_embed', 'cls_token', 'dist_token'}

    @torch.jit.ignore
    def group_matcher(self, coarse=False):
        return dict(
            stem=r'^cls_token|pos_embed|patch_embed',  # stem and embed
            blocks=[(r'^blocks\.(\d+)', None), (r'^norm', (99999,))]
        )

    @torch.jit.ignore
    def set_grad_checkpointing(self, enable=True):
        self.grad_checkpointing = enable

    @torch.jit.ignore
    def get_classifier(self):
        return self.head

    def reset_classifier(self, num_classes: int, global_pool=None, representation_size=None):
        self.num_classes = num_classes
        if global_pool is not None:
            assert global_pool in ('', 'avg', 'token')
            self.global_pool = global_pool
        if representation_size is not None:
            self._reset_representation(representation_size)
        final_chs = self.representation_size if self.representation_size else self.embed_dim
        self.head = nn.Linear(final_chs, num_classes) if num_classes > 0 else nn.Identity()

    def forward_features(self, x):
        x = self.patch_embed(x)
        x = torch.cat((self.cls_token.expand(x.shape[0], -1, -1), x), dim=1)

        x = self.pos_drop(x + self.pos_embed)
        if self.grad_checkpointing and not torch.jit.is_scripting():
            x = checkpoint_seq(self.blocks, x)
        else:
            x = self.blocks(x)
        x = self.norm(x)
        return x

    def forward_features_grow(self, x, class_num):
        x = self.patch_embed(x)
        # x = torch.cat((self.cls_token_grow[:, :class_num, :].expand(x.shape[0], -1, -1), self.cls_token.expand(x.shape[0], -1, -1), x), dim=1)
        # x = self.pos_drop(x + self.pos_embed_grow[:, :self.patch_embed.num_patches+class_num, :])
        x = torch.cat((self.cls_token.expand(x.shape[0], -1, -1), x), dim=1)
        x = self.pos_drop(x + self.pos_embed)
        x = torch.cat((self.cls_token_grow[:, :class_num*2, :].expand(x.shape[0], -1, -1), x), dim=1)

        # import pdb;pdb.set_trace()
        if self.grad_checkpointing and not torch.jit.is_scripting():
            x = checkpoint_seq(self.blocks, x)
        else:
            x = self.blocks(x)
        x = self.norm(x)
        return x

    def forward_head(self, x, pre_logits: bool = False):
        if self.global_pool:
            x = x[:, 1:].mean(dim=1) if self.global_pool == 'avg' else x[:, 0]
        x = self.fc_norm(x)
        x = self.pre_logits(x)
        return x if pre_logits else self.head(x)

    def forward(self, x, grow_flag=False, numcls=0):
        if not grow_flag:
            x = self.forward_features(x)
        else:
            x = self.forward_features_grow(x, numcls)

        if self.global_pool:
            x = x[:, 1:].mean(dim=1) if self.global_pool == 'avg' else x[:, 0]
        x = self.fc_norm(x)
        return {
            'fmaps': [x],
            'features': x
        }


def init_weights_vit_timm(module: nn.Module, name: str = ''):
    """ ViT weight initialization, original timm impl (for reproducibility) """
    if isinstance(module, nn.Linear):
        trunc_normal_(module.weight, std=.02)
        if module.bias is not None:
            nn.init.zeros_(module.bias)


def init_weights_vit_jax(module: nn.Module, name: str = '', head_bias: float = 0.):
    """ ViT weight initialization, matching JAX (Flax) impl """
    if isinstance(module, nn.Linear):
        if name.startswith('head'):
            nn.init.zeros_(module.weight)
            nn.init.constant_(module.bias, head_bias)
        elif name.startswith('pre_logits'):
            lecun_normal_(module.weight)
            nn.init.zeros_(module.bias)
        else:
            nn.init.xavier_uniform_(module.weight)
            if module.bias is not None:
                nn.init.normal_(module.bias, std=1e-6) if 'mlp' in name else nn.init.zeros_(module.bias)
    elif isinstance(module, nn.Conv2d):
        lecun_normal_(module.weight)
        if module.bias is not None:
            nn.init.zeros_(module.bias)


def init_weights_vit_moco(module: nn.Module, name: str = ''):
    """ ViT weight initialization, matching moco-v3 impl minus fixed PatchEmbed """
    if isinstance(module, nn.Linear):
        if 'qkv' in name:
            # treat the weights of Q, K, V separately
            val = math.sqrt(6. / float(module.weight.shape[0] // 3 + module.weight.shape[1]))
            nn.init.uniform_(module.weight, -val, val)
        else:
            nn.init.xavier_uniform_(module.weight)
        if module.bias is not None:
            nn.init.zeros_(module.bias)


def get_init_weights_vit(mode='jax', head_bias: float = 0.):
    if 'jax' in mode:
        return partial(init_weights_vit_jax, head_bias=head_bias)
    elif 'moco' in mode:
        return init_weights_vit_moco
    else:
        return init_weights_vit_timm


@torch.no_grad()
def _load_weights(model: VisionTransformer, checkpoint_path: str, prefix: str = ''):
    """ Load weights from .npz checkpoints for official Google Brain Flax implementation
    """
    import numpy as np

    def _n2p(w, t=True):
        if w.ndim == 4 and w.shape[0] == w.shape[1] == w.shape[2] == 1:
            w = w.flatten()
        if t:
            if w.ndim == 4:
                w = w.transpose([3, 2, 0, 1])
            elif w.ndim == 3:
                w = w.transpose([2, 0, 1])
            elif w.ndim == 2:
                w = w.transpose([1, 0])
        return torch.from_numpy(w)

    w = np.load(checkpoint_path)
    if not prefix and 'opt/target/embedding/kernel' in w:
        prefix = 'opt/target/'

    if hasattr(model.patch_embed, 'backbone'):
        # hybrid
        backbone = model.patch_embed.backbone
        stem_only = not hasattr(backbone, 'stem')
        stem = backbone if stem_only else backbone.stem
        stem.conv.weight.copy_(adapt_input_conv(stem.conv.weight.shape[1], _n2p(w[f'{prefix}conv_root/kernel'])))
        stem.norm.weight.copy_(_n2p(w[f'{prefix}gn_root/scale']))
        stem.norm.bias.copy_(_n2p(w[f'{prefix}gn_root/bias']))
        if not stem_only:
            for i, stage in enumerate(backbone.stages):
                for j, block in enumerate(stage.blocks):
                    bp = f'{prefix}block{i + 1}/unit{j + 1}/'
                    for r in range(3):
                        getattr(block, f'conv{r + 1}').weight.copy_(_n2p(w[f'{bp}conv{r + 1}/kernel']))
                        getattr(block, f'norm{r + 1}').weight.copy_(_n2p(w[f'{bp}gn{r + 1}/scale']))
                        getattr(block, f'norm{r + 1}').bias.copy_(_n2p(w[f'{bp}gn{r + 1}/bias']))
                    if block.downsample is not None:
                        block.downsample.conv.weight.copy_(_n2p(w[f'{bp}conv_proj/kernel']))
                        block.downsample.norm.weight.copy_(_n2p(w[f'{bp}gn_proj/scale']))
                        block.downsample.norm.bias.copy_(_n2p(w[f'{bp}gn_proj/bias']))
        embed_conv_w = _n2p(w[f'{prefix}embedding/kernel'])
    else:
        embed_conv_w = adapt_input_conv(
            model.patch_embed.proj.weight.shape[1], _n2p(w[f'{prefix}embedding/kernel']))
    model.patch_embed.proj.weight.copy_(embed_conv_w)
    model.patch_embed.proj.bias.copy_(_n2p(w[f'{prefix}embedding/bias']))
    model.cls_token.copy_(_n2p(w[f'{prefix}cls'], t=False))
    pos_embed_w = _n2p(w[f'{prefix}Transformer/posembed_input/pos_embedding'], t=False)
    if pos_embed_w.shape != model.pos_embed.shape:
        pos_embed_w = resize_pos_embed(  # resize pos embedding when different size from pretrained weights
            pos_embed_w, model.pos_embed, getattr(model, 'num_tokens', 1), model.patch_embed.grid_size)
    model.pos_embed.copy_(pos_embed_w)
    model.norm.weight.copy_(_n2p(w[f'{prefix}Transformer/encoder_norm/scale']))
    model.norm.bias.copy_(_n2p(w[f'{prefix}Transformer/encoder_norm/bias']))
    if isinstance(model.head, nn.Linear) and model.head.bias.shape[0] == w[f'{prefix}head/bias'].shape[-1]:
        model.head.weight.copy_(_n2p(w[f'{prefix}head/kernel']))
        model.head.bias.copy_(_n2p(w[f'{prefix}head/bias']))
    if isinstance(getattr(model.pre_logits, 'fc', None), nn.Linear) and f'{prefix}pre_logits/bias' in w:
        model.pre_logits.fc.weight.copy_(_n2p(w[f'{prefix}pre_logits/kernel']))
        model.pre_logits.fc.bias.copy_(_n2p(w[f'{prefix}pre_logits/bias']))
    for i, block in enumerate(model.blocks.children()):
        block_prefix = f'{prefix}Transformer/encoderblock_{i}/'
        mha_prefix = block_prefix + 'MultiHeadDotProductAttention_1/'
        block.norm1.weight.copy_(_n2p(w[f'{block_prefix}LayerNorm_0/scale']))
        block.norm1.bias.copy_(_n2p(w[f'{block_prefix}LayerNorm_0/bias']))
        block.attn.qkv.weight.copy_(torch.cat([
            _n2p(w[f'{mha_prefix}{n}/kernel'], t=False).flatten(1).T for n in ('query', 'key', 'value')]))
        block.attn.qkv.bias.copy_(torch.cat([
            _n2p(w[f'{mha_prefix}{n}/bias'], t=False).reshape(-1) for n in ('query', 'key', 'value')]))
        block.attn.proj.weight.copy_(_n2p(w[f'{mha_prefix}out/kernel']).flatten(1))
        block.attn.proj.bias.copy_(_n2p(w[f'{mha_prefix}out/bias']))
        for r in range(2):
            getattr(block.mlp, f'fc{r + 1}').weight.copy_(_n2p(w[f'{block_prefix}MlpBlock_3/Dense_{r}/kernel']))
            getattr(block.mlp, f'fc{r + 1}').bias.copy_(_n2p(w[f'{block_prefix}MlpBlock_3/Dense_{r}/bias']))
        block.norm2.weight.copy_(_n2p(w[f'{block_prefix}LayerNorm_2/scale']))
        block.norm2.bias.copy_(_n2p(w[f'{block_prefix}LayerNorm_2/bias']))


def resize_pos_embed(posemb, posemb_new, num_tokens=1, gs_new=()):
    # Rescale the grid of position embeddings when loading from state_dict. Adapted from
    # https://github.com/google-research/vision_transformer/blob/00883dd691c63a6830751563748663526e811cee/vit_jax/checkpoint.py#L224
    _logger.info('Resized position embedding: %s to %s', posemb.shape, posemb_new.shape)
    ntok_new = posemb_new.shape[1]
    if num_tokens:
        posemb_tok, posemb_grid = posemb[:, :num_tokens], posemb[0, num_tokens:]
        ntok_new -= num_tokens
    else:
        posemb_tok, posemb_grid = posemb[:, :0], posemb[0]
    gs_old = int(math.sqrt(len(posemb_grid)))
    if not len(gs_new):  # backwards compatibility
        gs_new = [int(math.sqrt(ntok_new))] * 2
    assert len(gs_new) >= 2
    _logger.info('Position embedding grid-size from %s to %s', [gs_old, gs_old], gs_new)
    posemb_grid = posemb_grid.reshape(1, gs_old, gs_old, -1).permute(0, 3, 1, 2)
    posemb_grid = F.interpolate(posemb_grid, size=gs_new, mode='bicubic', align_corners=False)
    posemb_grid = posemb_grid.permute(0, 2, 3, 1).reshape(1, gs_new[0] * gs_new[1], -1)
    posemb = torch.cat([posemb_tok, posemb_grid], dim=1)
    return posemb


def checkpoint_filter_fn(state_dict, model):
    """ convert patch embedding weight from manual patchify + linear proj to conv"""
    out_dict = {}
    if 'model' in state_dict:
        # For deit models
        state_dict = state_dict['model']
    for k, v in state_dict.items():
        if 'patch_embed.proj.weight' in k and len(v.shape) < 4:
            # For old models that I trained prior to conv based patchification
            O, I, H, W = model.patch_embed.proj.weight.shape
            v = v.reshape(O, -1, H, W)
        elif k == 'pos_embed' and v.shape != model.pos_embed.shape:
            # To resize pos embedding when using model at different size from pretrained weights
            v = resize_pos_embed(
                v, model.pos_embed, getattr(model, 'num_tokens', 1), model.patch_embed.grid_size)
        out_dict[k] = v
    return out_dict


def _create_vision_transformer(variant, pretrained=False, **kwargs):
    if kwargs.get('features_only', None):
        raise RuntimeError('features_only not implemented for Vision Transformer models.')

    # NOTE this extra code to support handling of repr size for in21k pretrained models
    # pretrained_cfg = resolve_pretrained_cfg(variant, kwargs=kwargs)
    pretrained_cfg = resolve_pretrained_cfg(variant)
    default_num_classes = pretrained_cfg['num_classes']
    num_classes = kwargs.get('num_classes', default_num_classes)
    repr_size = kwargs.pop('representation_size', None)
    if repr_size is not None and num_classes != default_num_classes:
        # Remove representation layer if fine-tuning. This may not always be the desired action,
        # but I feel better than doing nothing by default for fine-tuning. Perhaps a better interface?
        _logger.warning("Removing representation layer for fine-tuning.")
        repr_size = None

    if pretrained_cfg:
        del kwargs['pretrained_cfg']

    model = build_model_with_cfg(
        VisionTransformer, variant, pretrained,
        pretrained_cfg=pretrained_cfg,
        representation_size=repr_size,
        pretrained_filter_fn=checkpoint_filter_fn,
        pretrained_custom_load='npz' in pretrained_cfg['url'],
        **kwargs)
    return model