modeling_finetune.py

from functools import partial
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
from timm.models.layers import drop_path, to_2tuple, trunc_normal_
from timm.models.registry import register_model
from einops import rearrange, repeat


def _cfg(url='', **kwargs):
    return {
        'url': url,
        'num_classes': 400, 'input_size': (3, 224, 224), 'pool_size': None,
        'crop_pct': .9, 'interpolation': 'bicubic',
        'mean': (0.5, 0.5, 0.5), 'std': (0.5, 0.5, 0.5),
        **kwargs
    }


class DropPath(nn.Module):
    """Drop paths (Stochastic Depth) per sample  (when applied in main path of residual blocks).
    """
    def __init__(self, drop_prob=None):
        super(DropPath, self).__init__()
        self.drop_prob = drop_prob

    def forward(self, x):
        return drop_path(x, self.drop_prob, self.training)
    
    def extra_repr(self) -> str:
        return 'p={}'.format(self.drop_prob)


class Mlp(nn.Module):
    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Linear(in_features, hidden_features)
        self.act = act_layer()
        self.fc2 = nn.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)
        # commit this for the orignal BERT implement 
        x = self.fc2(x)
        x = self.drop(x)
        return x


class Attention(nn.Module):
    def __init__(
            self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0.,
            proj_drop=0., attn_head_dim=None):
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // num_heads
        if attn_head_dim is not None:
            head_dim = attn_head_dim
        all_head_dim = head_dim * self.num_heads
        self.scale = qk_scale or head_dim ** -0.5

        self.qkv = nn.Linear(dim, all_head_dim * 3, bias=False)
        if qkv_bias:
            self.q_bias = nn.Parameter(torch.zeros(all_head_dim))
            self.v_bias = nn.Parameter(torch.zeros(all_head_dim))
        else:
            self.q_bias = None
            self.v_bias = None

        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(all_head_dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

    def forward(self, x, mask=None):
        B, N, C = x.shape
        qkv_bias = None
        if self.q_bias is not None:
            qkv_bias = torch.cat((self.q_bias, torch.zeros_like(self.v_bias, requires_grad=False), self.v_bias))
        # qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        qkv = F.linear(input=x, weight=self.qkv.weight, bias=qkv_bias)
        qkv = qkv.reshape(B, N, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
        q, k, v = qkv[0], qkv[1], qkv[2]   # make torchscript happy (cannot use tensor as tuple)

        q = q * self.scale
        attn = (q @ k.transpose(-2, -1))

        # me: support window mask
        if mask is not None:
            nW = mask.shape[0]
            attn = attn.view(B // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0)
            attn = attn.view(-1, self.num_heads, N, N)
            attn = attn.softmax(dim=-1)
        else:
            attn = attn.softmax(dim=-1)

        attn = self.attn_drop(attn)

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


class Block(nn.Module):

    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,
                 drop_path=0., init_values=None, act_layer=nn.GELU, norm_layer=nn.LayerNorm,
                 attn_head_dim=None):
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention(
            dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale,
            attn_drop=attn_drop, proj_drop=drop, attn_head_dim=attn_head_dim)
        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
        self.drop_path = 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)

        if init_values > 0:
            self.gamma_1 = nn.Parameter(init_values * torch.ones((dim)),requires_grad=True)
            self.gamma_2 = nn.Parameter(init_values * torch.ones((dim)),requires_grad=True)
        else:
            self.gamma_1, self.gamma_2 = None, None

    def forward(self, x, mask=None):
        if self.gamma_1 is None:
            x = x + self.drop_path(self.attn(self.norm1(x), mask=mask))
            x = x + self.drop_path(self.mlp(self.norm2(x)))
        else:
            x = x + self.drop_path(self.gamma_1 * self.attn(self.norm1(x), mask=mask))
            x = x + self.drop_path(self.gamma_2 * self.mlp(self.norm2(x)))
        return x



"""
adapted from https://github.com/lucidrains/perceiver-pytorch/blob/main/perceiver_pytorch/perceiver_pytorch.py
"""
# support cross attention
class GeneralAttention(nn.Module):
    def __init__(
            self, dim, context_dim=None, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0.,
            proj_drop=0., attn_head_dim=None):
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // num_heads
        if attn_head_dim is not None:
            head_dim = attn_head_dim
        all_head_dim = head_dim * self.num_heads
        self.scale = qk_scale or head_dim ** -0.5

        self.q = nn.Linear(dim, all_head_dim, bias=False)
        self.kv = nn.Linear(dim if context_dim is None else context_dim, all_head_dim * 2, bias=False)
        if qkv_bias:
            self.q_bias = nn.Parameter(torch.zeros(all_head_dim))
            self.v_bias = nn.Parameter(torch.zeros(all_head_dim))
        else:
            self.q_bias = None
            self.v_bias = None

        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(all_head_dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

    def forward(self, x, context=None):
        B, T1, C = x.shape
        q_bias, kv_bias = self.q_bias, None
        if self.q_bias is not None:
            kv_bias = torch.cat((torch.zeros_like(self.v_bias, requires_grad=False), self.v_bias))
        q = F.linear(input=x, weight=self.q.weight, bias=q_bias)
        q = q.reshape(B, T1, self.num_heads, -1).transpose(1,2) # me: (B, H, T1, C//H)
        kv = F.linear(input=x if context is None else context, weight=self.kv.weight, bias=kv_bias)
        _, T2, _ = kv.shape
        kv = kv.reshape(B, T2, 2, self.num_heads, -1).permute(2, 0, 3, 1, 4)
        k, v = kv[0], kv[1] # make torchscript happy (cannot use tensor as tuple), me： (B, H, T2, C//H)

        q = q * self.scale
        attn = (q @ k.transpose(-2, -1)) # me: (B, H, T1, T2)

        attn = attn.softmax(dim=-1)
        attn = self.attn_drop(attn)

        x = (attn @ v).transpose(1, 2).reshape(B, T1, -1) # (B, H, T1, C//H) -> (B, T1, H, C//H) -> (B, T1, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x



# local + global
class LGBlock(nn.Module):
    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,
                 drop_path=0., init_values=None, act_layer=nn.GELU, norm_layer=nn.LayerNorm,
                 attn_head_dim=None,
                 # new added
                 first_attn_type='self', third_attn_type='cross',
                 attn_param_sharing_first_third=False, attn_param_sharing_all=False,
                 no_second=False, no_third=False,
                 ):

        super().__init__()

        assert first_attn_type in ['self', 'cross'], f"Error: invalid attention type '{first_attn_type}', expected 'self' or 'cross'!"
        assert third_attn_type in ['self', 'cross'], f"Error: invalid attention type '{third_attn_type}', expected 'self' or 'cross'!"
        self.first_attn_type = first_attn_type
        self.third_attn_type = third_attn_type
        self.attn_param_sharing_first_third = attn_param_sharing_first_third
        self.attn_param_sharing_all = attn_param_sharing_all

        # Attention layer
        ## perform local (intra-region) attention, update messenger tokens
        ## (local->messenger) or (local<->local, local<->messenger)
        self.first_attn_norm0 = norm_layer(dim)
        if self.first_attn_type == 'cross':
            self.first_attn_norm1 = norm_layer(dim)
        self.first_attn = GeneralAttention(
            dim=dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale,
            attn_drop=attn_drop, proj_drop=drop, attn_head_dim=attn_head_dim)

        ## perform global (inter-region) attention on messenger tokens
        ## (messenger<->messenger)
        self.no_second = no_second
        if not no_second:
            self.second_attn_norm0 = norm_layer(dim)
            if attn_param_sharing_all:
                self.second_attn = self.first_attn
            else:
                self.second_attn = GeneralAttention(
                    dim=dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale,
                    attn_drop=attn_drop, proj_drop=drop, attn_head_dim=attn_head_dim)

        ## perform local (intra-region) attention to inject global information into local tokens
        ## (messenger->local) or (local<->local, local<->messenger)
        self.no_third = no_third
        if not no_third:
            self.third_attn_norm0 = norm_layer(dim)
            if self.third_attn_type == 'cross':
                self.third_attn_norm1 = norm_layer(dim)
            if attn_param_sharing_first_third or attn_param_sharing_all:
                self.third_attn = self.first_attn
            else:
                self.third_attn = GeneralAttention(
                    dim=dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale,
                    attn_drop=attn_drop, proj_drop=drop, attn_head_dim=attn_head_dim)

        # FFN layer
        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
        self.drop_path = 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)

        if init_values > 0:
            self.gamma_0 = nn.Parameter(init_values * torch.ones((dim)),requires_grad=True)
            self.gamma_1 = nn.Parameter(init_values * torch.ones((dim)),requires_grad=True)
            self.gamma_2 = nn.Parameter(init_values * torch.ones((dim)),requires_grad=True)
        else:
            self.gamma_0, self.gamma_1, self.gamma_2 = None, None, None


    def forward(self, x, b):
        """
        :param x: (B*N, S, C),
            B: batch size
            N: number of local regions
            S: 1 + region size, 1: attached messenger token for each local region
            C: feature dim
        param b: batch size
        :return: (B*N, S, C),
        """
        bn = x.shape[0]
        n = bn // b # number of local regions
        if self.gamma_1 is None:
            # Attention layer
            ## perform local (intra-region) self-attention
            if self.first_attn_type == 'self':
                x = x + self.drop_path(self.first_attn(self.first_attn_norm0(x)))
            else: # 'cross'
                x[:,:1] = x[:,:1] + self.drop_path(
                    self.first_attn(
                        self.first_attn_norm0(x[:,:1]), # (b*n, 1, c)
                        context=self.first_attn_norm1(x[:,1:]) # (b*n, s-1, c)
                    )
                )

            ## perform global (inter-region) self-attention
            if not self.no_second:
                # messenger_tokens: representative tokens
                messenger_tokens = rearrange(x[:,0], '(b n) c -> b n c', b=b) # attn on 'n' dim
                messenger_tokens = messenger_tokens + self.drop_path(
                    self.second_attn(self.second_attn_norm0(messenger_tokens))
                )
                x[:,0] = rearrange(messenger_tokens, 'b n c -> (b n) c')
            else: # for usage in the third attn
                messenger_tokens = rearrange(x[:,0], '(b n) c -> b n c', b=b) # attn on 'n' dim

            ## perform local-global interaction
            if not self.no_third:
                if self.third_attn_type == 'self':
                    x = x + self.drop_path(self.third_attn(self.third_attn_norm0(x)))
                else:
                    local_tokens = rearrange(x[:,1:], '(b n) s c -> b (n s) c', b=b)# NOTE: n merges into s (not b), (B, N*(S-1), D)
                    local_tokens = local_tokens + self.drop_path(
                        self.third_attn(
                            self.third_attn_norm0(local_tokens), # (b, n*(s-1), c)
                            context=self.third_attn_norm1(messenger_tokens) # (b, n*1, c)
                        )
                    )
                    x[:,1:] = rearrange(local_tokens, 'b (n s) c -> (b n) s c', n=n)

            # FFN layer
            x = x + self.drop_path(self.mlp(self.norm2(x)))
        else:
            raise NotImplementedError
        return x


class PatchEmbed(nn.Module):
    """ Image to Patch Embedding
    """
    def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768, num_frames=16, tubelet_size=2):
        super().__init__()
        img_size = to_2tuple(img_size)
        patch_size = to_2tuple(patch_size)
        self.tubelet_size = int(tubelet_size)
        num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0]) * (num_frames // self.tubelet_size)
        self.img_size = img_size
        self.patch_size = patch_size
        self.num_patches = num_patches

        # me: for more attention types
        self.temporal_seq_len = num_frames // self.tubelet_size
        self.spatial_num_patches = num_patches // self.temporal_seq_len
        self.input_token_size = (num_frames // self.tubelet_size, img_size[0] // patch_size[0], img_size[1] // patch_size[1])

        self.proj = nn.Conv3d(in_channels=in_chans, out_channels=embed_dim, 
                            kernel_size = (self.tubelet_size,  patch_size[0],patch_size[1]), 
                            stride=(self.tubelet_size,  patch_size[0],  patch_size[1]))

    def forward(self, x, **kwargs):
        B, C, T, H, W = x.shape
        # FIXME look at relaxing size constraints
        assert H == self.img_size[0] and W == self.img_size[1], \
            f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})."
        x = self.proj(x).flatten(2).transpose(1, 2)
        return x

# sin-cos position encoding
# https://github.com/jadore801120/attention-is-all-you-need-pytorch/blob/master/transformer/Models.py#L31
def get_sinusoid_encoding_table(n_position, d_hid): 
    ''' Sinusoid position encoding table ''' 
    # TODO: make it with torch instead of numpy 
    def get_position_angle_vec(position): 
        return [position / np.power(10000, 2 * (hid_j // 2) / d_hid) for hid_j in range(d_hid)] 

    sinusoid_table = np.array([get_position_angle_vec(pos_i) for pos_i in range(n_position)]) 
    sinusoid_table[:, 0::2] = np.sin(sinusoid_table[:, 0::2]) # dim 2i 
    sinusoid_table[:, 1::2] = np.cos(sinusoid_table[:, 1::2]) # dim 2i+1 

    return torch.FloatTensor(sinusoid_table).unsqueeze(0) 


class VisionTransformer(nn.Module):
    """ Vision Transformer with support for patch or hybrid CNN input stage
    """
    def __init__(self, 
                 img_size=224, 
                 patch_size=16, 
                 in_chans=3, 
                 num_classes=1000, 
                 embed_dim=768, 
                 depth=12,
                 num_heads=12, 
                 mlp_ratio=4., 
                 qkv_bias=False, 
                 qk_scale=None, 
                 drop_rate=0., 
                 attn_drop_rate=0.,
                 drop_path_rate=0., 
                 norm_layer=nn.LayerNorm, 
                 init_values=0.,
                 use_learnable_pos_emb=False, 
                 init_scale=0.,
                 all_frames=16,
                 tubelet_size=2,
                 use_mean_pooling=True,
                 keep_temporal_dim=False, # do not perform temporal pooling, has higher priority than 'use_mean_pooling'
                 head_activation_func=None, # activation function after head fc, mainly for the regression task
                 attn_type='joint',
                 lg_region_size=(2, 2, 10), lg_first_attn_type='self', lg_third_attn_type='cross',  # for local_global
                 lg_attn_param_sharing_first_third=False, lg_attn_param_sharing_all=False,
                 lg_classify_token_type='org', lg_no_second=False, lg_no_third=False,
                 ):
        super().__init__()
        self.num_classes = num_classes
        self.num_features = self.embed_dim = embed_dim  # num_features for consistency with other models
        self.tubelet_size = tubelet_size
        self.patch_embed = PatchEmbed(
            img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim, num_frames=all_frames, tubelet_size=self.tubelet_size)
        num_patches = self.patch_embed.num_patches

        if use_learnable_pos_emb:
            self.pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim))
        else:
            # sine-cosine positional embeddings is on the way
            self.pos_embed = get_sinusoid_encoding_table(num_patches, 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
        # me: support more attention types
        self.attn_type = attn_type
        if attn_type == 'joint':
            self.blocks = nn.ModuleList([
                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, drop_path=dpr[i], norm_layer=norm_layer,
                    init_values=init_values)
                for i in range(depth)])
        elif attn_type == 'local_global':
            print(f"==> Note: Use 'local_global' for compute reduction (lg_region_size={lg_region_size},"
                  f"lg_first_attn_type={lg_first_attn_type}, lg_third_attn_type={lg_third_attn_type},"
                  f"lg_attn_param_sharing_first_third={lg_attn_param_sharing_first_third},"
                  f"lg_attn_param_sharing_all={lg_attn_param_sharing_all},"
                  f"lg_classify_token_type={lg_classify_token_type},"
                  f"lg_no_second={lg_no_second}, lg_no_third={lg_no_third})")
            self.blocks = nn.ModuleList([
                LGBlock(
                    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, drop_path=dpr[i], norm_layer=norm_layer,
                    init_values=init_values,
                    first_attn_type=lg_first_attn_type, third_attn_type=lg_third_attn_type,
                    attn_param_sharing_first_third=lg_attn_param_sharing_first_third,
                    attn_param_sharing_all=lg_attn_param_sharing_all,
                    no_second=lg_no_second, no_third=lg_no_third,
                )
                for i in range(depth)])
            # region tokens
            self.lg_region_size = lg_region_size # (t, h, w)
            self.lg_num_region_size = list(i//j for i,j in zip(self.patch_embed.input_token_size, lg_region_size)) # (nt, nh, nw)
            num_regions = self.lg_num_region_size[0] * self.lg_num_region_size[1] * self.lg_num_region_size[2] # nt * nh * nw
            print(f"==> Number of local regions: {num_regions} (size={self.lg_num_region_size})")
            self.lg_region_tokens = nn.Parameter(torch.zeros(num_regions, embed_dim))
            trunc_normal_(self.lg_region_tokens, std=.02)

            # The token type used for final classification
            self.lg_classify_token_type = lg_classify_token_type
            assert lg_classify_token_type in ['org', 'region', 'all'], \
                f"Error: wrong 'lg_classify_token_type' in local_global attention ('{lg_classify_token_type}'), expected 'org'/'region'/'all'!"

        else:
            raise NotImplementedError

        self.norm = nn.Identity() if use_mean_pooling else norm_layer(embed_dim)
        self.fc_norm = norm_layer(embed_dim) if use_mean_pooling else None
        self.head = nn.Linear(embed_dim, num_classes) if num_classes > 0 else nn.Identity()

        # me: add frame-level prediction support
        self.keep_temporal_dim = keep_temporal_dim

        # me: add head activation function support for regression task
        if head_activation_func is not None:
            if head_activation_func == 'sigmoid':
                self.head_activation_func = nn.Sigmoid()
            elif head_activation_func == 'relu':
                self.head_activation_func = nn.ReLU()
            elif head_activation_func == 'tanh':
                self.head_activation_func = nn.Tanh()
            else:
                raise NotImplementedError
        else: # default
            self.head_activation_func = nn.Identity()

        if use_learnable_pos_emb:
            trunc_normal_(self.pos_embed, std=.02)

        trunc_normal_(self.head.weight, std=.02)
        self.apply(self._init_weights)

        self.head.weight.data.mul_(init_scale)
        self.head.bias.data.mul_(init_scale)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)

    def get_num_layers(self):
        return len(self.blocks)

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

    def get_classifier(self):
        return self.head

    def reset_classifier(self, num_classes, global_pool=''):
        self.num_classes = num_classes
        self.head = nn.Linear(self.embed_dim, num_classes) if num_classes > 0 else nn.Identity()

    def forward_features(self, x):
        x = self.patch_embed(x)
        B, _, _ = x.size()

        if self.pos_embed is not None:
            x = x + self.pos_embed.expand(B, -1, -1).type_as(x).to(x.device).clone().detach()
        x = self.pos_drop(x)

        if self.attn_type == 'local_global':
            # input: region partition
            nt, t = self.lg_num_region_size[0], self.lg_region_size[0]
            nh, h = self.lg_num_region_size[1], self.lg_region_size[1]
            nw, w = self.lg_num_region_size[2], self.lg_region_size[2]
            b = x.size(0)
            x = rearrange(x, 'b (nt t nh h nw w) c -> b (nt nh nw) (t h w) c', nt=nt,nh=nh,nw=nw,t=t,h=h,w=w)
            # add region (i.e., representative) tokens
            region_tokens = repeat(self.lg_region_tokens, 'n c -> b n 1 c', b=b)
            x = torch.cat([region_tokens, x], dim=2) # (b, nt*nh*nw, 1+thw, c)
            x = rearrange(x, 'b n s c -> (b n) s c') # s = 1 + thw
            # run through each block
            for blk in self.blocks:
                x = blk(x, b) # (b*n, s, c)

            x = rearrange(x, '(b n) s c -> b n s c', b=b) # s = 1 + thw
            # token for final classification
            if self.lg_classify_token_type == 'region': # only use region tokens for classification
                x = x[:,:,0] # (b, n, c)
            elif self.lg_classify_token_type == 'org': # only use original tokens for classification
                x = rearrange(x[:,:,1:], 'b n s c -> b (n s) c') # s = thw
            else: # use all tokens for classification
                x = rearrange(x, 'b n s c -> b (n s) c') # s = 1 + thw

        else:
            for blk in self.blocks:
                x = blk(x)

        x = self.norm(x)
        if self.fc_norm is not None:
            # me: add frame-level prediction support
            if self.keep_temporal_dim:
                x = rearrange(x, 'b (t hw) c -> b c t hw',
                              t=self.patch_embed.temporal_seq_len,
                              hw=self.patch_embed.spatial_num_patches)
                # spatial mean pooling
                x = x.mean(-1) # (B, C, T)
                # temporal upsample: 8 -> 16, for patch embedding reduction
                x = torch.nn.functional.interpolate(
                    x, scale_factor=self.patch_embed.tubelet_size,
                    mode='linear'
                )
                x = rearrange(x, 'b c t -> b t c')
                return self.fc_norm(x)
            else:
                return self.fc_norm(x.mean(1))
        else:
            return x[:, 0]

    def forward(self, x, save_feature=False):
        x = self.forward_features(x)
        if save_feature:
            feature = x
        x = self.head(x)
        # me: add head activation function support
        x = self.head_activation_func(x)
        # me: add frame-level prediction support
        if self.keep_temporal_dim:
            x = x.view(x.size(0), -1) # (B,T,C) -> (B,T*C)
        if save_feature:
            return x, feature
        else:
            return x




@register_model
def vit_base_patch16_160(pretrained=False, **kwargs):
    model = VisionTransformer(
        img_size=160,
        patch_size=16, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


@register_model
def vit_base_dim512_no_depth_patch16_160(pretrained=False, **kwargs):
    model = VisionTransformer(
        img_size=160,
        patch_size=16, embed_dim=512, num_heads=8, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model