Overhaul FocalNet implementation

timm/models/focalnet.py

@@ -1,19 +1,32 @@
+""" FocalNet
+
+As described in `Focal Modulation Networks` - https://arxiv.org/abs/2203.11926
+
+Significant modifications and refactoring from the original impl at https://github.com/microsoft/FocalNet
+
+This impl is/has:
+* fully convolutional, NCHW tensor layout throughout, seemed to have minimal performance impact but more flexible
+* re-ordered downsample / layer so that striding always at beginning of layer (stage)
+* no input size constraints or input resolution/H/W tracking through the model
+* torchscript fixed and a number of quirks cleaned up
+* feature extraction support via `features_only=True`
+"""
 # --------------------------------------------------------
 # FocalNets -- Focal Modulation Networks
 # Copyright (c) 2022 Microsoft
 # Licensed under The MIT License [see LICENSE for details]
 # Written by Jianwei Yang (jianwyan@microsoft.com)
 # --------------------------------------------------------
+from functools import partial
 
 import torch
 import torch.nn as nn
-import torch.nn.functional as F
 import torch.utils.checkpoint as checkpoint
 
 from timm.data import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD
-from timm.layers import PatchEmbed, Mlp, DropPath, to_2tuple, trunc_normal_, _assert
+from timm.layers import Mlp, DropPath, LayerNorm2d, trunc_normal_, ClassifierHead, NormMlpClassifierHead
 from ._builder import build_model_with_cfg
-from ._features_fx import register_notrace_function
+from ._manipulate import named_apply
 from ._registry import register_model
 
 __all__ = ['FocalNet']
@@ -27,9 +40,10 @@ class FocalModulation(nn.Module):
             focal_level,
             focal_factor=2,
             bias=True,
-            proj_drop=0.,
-            use_postln_in_modulation=False,
+            use_post_norm=False,
             normalize_modulator=False,
+            proj_drop=0.,
+            norm_layer=LayerNorm2d,
     ):
         super().__init__()
 
@@ -37,69 +51,70 @@ class FocalModulation(nn.Module):
         self.focal_window = focal_window
         self.focal_level = focal_level
         self.focal_factor = focal_factor
-        self.use_postln_in_modulation = use_postln_in_modulation
+        self.use_post_norm = use_post_norm
         self.normalize_modulator = normalize_modulator
+        self.input_split = [dim, dim, self.focal_level + 1]
 
-        self.f = nn.Linear(dim, 2 * dim + (self.focal_level + 1), bias=bias)
-        self.h = nn.Conv2d(dim, dim, kernel_size=1, stride=1, bias=bias)
+        self.f = nn.Conv2d(dim, 2 * dim + (self.focal_level + 1), kernel_size=1, bias=bias)
+        self.h = nn.Conv2d(dim, dim, kernel_size=1, bias=bias)
 
         self.act = nn.GELU()
-        self.proj = nn.Linear(dim, dim)
+        self.proj = nn.Conv2d(dim, dim, kernel_size=1)
         self.proj_drop = nn.Dropout(proj_drop)
         self.focal_layers = nn.ModuleList()
 
         self.kernel_sizes = []
         for k in range(self.focal_level):
             kernel_size = self.focal_factor * k + self.focal_window
-            self.focal_layers.append(
-                nn.Sequential(
-                    nn.Conv2d(
-                        dim, dim, kernel_size=kernel_size, stride=1,
-                        groups=dim, padding=kernel_size // 2, bias=False),
-                    nn.GELU(),
-                )
-            )
+            self.focal_layers.append(nn.Sequential(
+                nn.Conv2d(dim, dim, kernel_size=kernel_size, groups=dim, padding=kernel_size // 2, bias=False),
+                nn.GELU(),
+            ))
             self.kernel_sizes.append(kernel_size)
-        if self.use_postln_in_modulation:
-            self.ln = nn.LayerNorm(dim)
+        self.norm = norm_layer(dim) if self.use_post_norm else nn.Identity()
 
     def forward(self, x):
         """
         Args:
             x: input features with shape of (B, H, W, C)
         """
-        C = x.shape[-1]
+        C = x.shape[1]
 
         # pre linear projection
-        x = self.f(x).permute(0, 3, 1, 2).contiguous()
-        q, ctx, self.gates = torch.split(x, (C, C, self.focal_level + 1), 1)
+        x = self.f(x)
+        q, ctx, gates = torch.split(x, self.input_split, 1)
 
         # context aggreation
         ctx_all = 0
-        for l in range(self.focal_level):
-            ctx = self.focal_layers[l](ctx)
-            ctx_all = ctx_all + ctx * self.gates[:, l:l + 1]
-        ctx_global = self.act(ctx.mean(2, keepdim=True).mean(3, keepdim=True))
-        ctx_all = ctx_all + ctx_global * self.gates[:, self.focal_level:]
+        for l, focal_layer in enumerate(self.focal_layers):
+            ctx = focal_layer(ctx)
+            ctx_all = ctx_all + ctx * gates[:, l:l + 1]
+        ctx_global = self.act(ctx.mean((2, 3), keepdim=True))
+        ctx_all = ctx_all + ctx_global * gates[:, self.focal_level:]
 
         # normalize context
         if self.normalize_modulator:
             ctx_all = ctx_all / (self.focal_level + 1)
 
         # focal modulation
-        self.modulator = self.h(ctx_all)
-        x_out = q * self.modulator
-        x_out = x_out.permute(0, 2, 3, 1).contiguous()
-        if self.use_postln_in_modulation:
-            x_out = self.ln(x_out)
+        x_out = q * self.h(ctx_all)
+        x_out = self.norm(x_out)
 
-        # post linear porjection
+        # post linear projection
         x_out = self.proj(x_out)
         x_out = self.proj_drop(x_out)
         return x_out
 
-    def extra_repr(self) -> str:
-        return f'dim={self.dim}'
+
+class LayerScale2d(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):
+        gamma = self.gamma.view(1, -1, 1, 1)
+        return x.mul_(gamma) if self.inplace else x * gamma
 
 
 class FocalNetBlock(nn.Module):
@@ -107,297 +122,238 @@ class FocalNetBlock(nn.Module):
 
     Args:
         dim (int): Number of input channels.
-        input_resolution (tuple[int]): Input resulotion.
         mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
-        drop (float, optional): Dropout rate. Default: 0.0
+        proj_drop (float, optional): Dropout rate. Default: 0.0
         drop_path (float, optional): Stochastic depth rate. Default: 0.0
         act_layer (nn.Module, optional): Activation layer. Default: nn.GELU
         norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
         focal_level (int): Number of focal levels.
         focal_window (int): Focal window size at first focal level
         layerscale_value (float): Initial layerscale value
-        use_postln (bool): Whether to use layernorm after modulation
+        use_post_norm (bool): Whether to use layernorm after modulation
     """
 
     def __init__(
             self,
             dim,
-            input_resolution,
             mlp_ratio=4.,
-            drop=0.,
-            drop_path=0.,
-            act_layer=nn.GELU,
-            norm_layer=nn.LayerNorm,
             focal_level=1,
             focal_window=3,
-            layerscale_value=1e-4,
-            use_postln=False,
-            use_postln_in_modulation=False,
+            use_post_norm=False,
+            use_post_norm_in_modulation=False,
             normalize_modulator=False,
+            layerscale_value=1e-4,
+            proj_drop=0.,
+            drop_path=0.,
+            act_layer=nn.GELU,
+            norm_layer=LayerNorm2d,
     ):
         super().__init__()
         self.dim = dim
-        self.input_resolution = input_resolution
         self.mlp_ratio = mlp_ratio
 
         self.focal_window = focal_window
         self.focal_level = focal_level
-        self.use_postln = use_postln
+        self.use_post_norm = use_post_norm
 
-        self.norm1 = norm_layer(dim)
+        self.norm1 = norm_layer(dim) if not use_post_norm else nn.Identity()
         self.modulation = FocalModulation(
             dim,
-            proj_drop=drop,
             focal_window=focal_window,
             focal_level=self.focal_level,
-            use_postln_in_modulation=use_postln_in_modulation,
+            use_post_norm=use_post_norm_in_modulation,
             normalize_modulator=normalize_modulator,
+            proj_drop=proj_drop,
+            norm_layer=norm_layer,
         )
+        self.norm1_post = norm_layer(dim) if use_post_norm else nn.Identity()
+        self.ls1 = LayerScale2d(dim, layerscale_value) if layerscale_value is not None else nn.Identity()
+        self.drop_path1 = DropPath(drop_path) if drop_path > 0. else nn.Identity()
 
-        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.norm2 = norm_layer(dim) if not use_post_norm else nn.Identity()
         self.mlp = Mlp(
             in_features=dim,
-            hidden_features=mlp_hidden_dim,
+            hidden_features=int(dim * mlp_ratio),
             act_layer=act_layer,
-            drop=drop,
+            drop=proj_drop,
+            use_conv=True,
         )
-
-        self.gamma_1 = 1.0
-        self.gamma_2 = 1.0
-        if layerscale_value is not None:
-            self.gamma_1 = nn.Parameter(layerscale_value * torch.ones(dim))
-            self.gamma_2 = nn.Parameter(layerscale_value * torch.ones(dim))
-
-        self.H = None
-        self.W = None
+        self.norm2_post = norm_layer(dim) if use_post_norm else nn.Identity()
+        self.ls2 = LayerScale2d(dim, layerscale_value) if layerscale_value is not None else nn.Identity()
+        self.drop_path2 = DropPath(drop_path) if drop_path > 0. else nn.Identity()
 
     def forward(self, x):
-        H, W = self.H, self.W
-        B, L, C = x.shape
         shortcut = x
 
         # Focal Modulation
-        x = x if self.use_postln else self.norm1(x)
-        x = x.view(B, H, W, C)
-        x = self.modulation(x).view(B, H * W, C)
-        x = x if not self.use_postln else self.norm1(x)
+        x = self.norm1(x)
+        x = self.modulation(x)
+        x = self.norm1_post(x)
+        x = shortcut + self.drop_path1(self.ls1(x))
 
         # FFN
-        x = shortcut + self.drop_path(self.gamma_1 * x)
-        x = x + self.drop_path(self.gamma_2 * (self.norm2(self.mlp(x)) if self.use_postln else self.mlp(self.norm2(x))))
+        x = x + self.drop_path2(self.ls2(self.norm2_post(self.mlp(self.norm2(x)))))
 
         return x
 
-    def extra_repr(self) -> str:
-        return f"dim={self.dim}, input_resolution={self.input_resolution}, " \
-               f"mlp_ratio={self.mlp_ratio}"
-
 
 class BasicLayer(nn.Module):
     """ A basic Focal Transformer layer for one stage.
 
     Args:
         dim (int): Number of input channels.
-        input_resolution (tuple[int]): Input resolution.
         depth (int): Number of blocks.
         mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
         drop (float, optional): Dropout rate. Default: 0.0
         drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0
         norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
-        downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None
-        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False.
+        downsample (bool): Downsample layer at start of the layer. Default: True
         focal_level (int): Number of focal levels
         focal_window (int): Focal window size at first focal level
         layerscale_value (float): Initial layerscale value
-        use_postln (bool): Whether to use layer norm after modulation
+        use_post_norm (bool): Whether to use layer norm after modulation
     """
 
     def __init__(
             self,
             dim,
             out_dim,
-            input_resolution,
             depth,
             mlp_ratio=4.,
-            drop=0.,
-            drop_path=0.,
-            norm_layer=nn.LayerNorm,
-            downsample=None,
-            use_checkpoint=False,
+            downsample=True,
             focal_level=1,
             focal_window=1,
-            use_conv_embed=False,
+            use_overlap_down=False,
+            use_post_norm=False,
+            use_post_norm_in_modulation=False,
+            normalize_modulator=False,
             layerscale_value=1e-4,
-            use_postln=False,
-            use_postln_in_modulation=False,
-            normalize_modulator=False
+            proj_drop=0.,
+            drop_path=0.,
+            norm_layer=LayerNorm2d,
     ):
 
         super().__init__()
         self.dim = dim
-        self.input_resolution = input_resolution
         self.depth = depth
-        self.use_checkpoint = use_checkpoint
+        self.grad_checkpointing = False
+
+        if downsample:
+            self.downsample = Downsample(
+                in_chs=dim,
+                out_chs=out_dim,
+                stride=2,
+                overlap=use_overlap_down,
+                norm_layer=norm_layer,
+            )
+        else:
+            self.downsample = nn.Identity()
 
         # build blocks
         self.blocks = nn.ModuleList([
             FocalNetBlock(
-                dim=dim,
-                input_resolution=input_resolution,
+                dim=out_dim,
                 mlp_ratio=mlp_ratio,
-                drop=drop,
-                drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path,
-                norm_layer=norm_layer,
                 focal_level=focal_level,
                 focal_window=focal_window,
-                layerscale_value=layerscale_value,
-                use_postln=use_postln,
-                use_postln_in_modulation=use_postln_in_modulation,
+                use_post_norm=use_post_norm,
+                use_post_norm_in_modulation=use_post_norm_in_modulation,
                 normalize_modulator=normalize_modulator,
+                layerscale_value=layerscale_value,
+                proj_drop=proj_drop,
+                drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path,
+                norm_layer=norm_layer,
             )
             for i in range(depth)])
 
-        if downsample is not None:
-            self.downsample = downsample(
-                img_size=input_resolution,
-                patch_size=2,
-                in_chans=dim,
-                embed_dim=out_dim,
-                use_conv_embed=use_conv_embed,
-                norm_layer=norm_layer,
-                is_stem=False
-            )
-        else:
-            self.downsample = None
-
-    def forward(self, x, H, W):
+    def forward(self, x):
+        x = self.downsample(x)
         for blk in self.blocks:
-            blk.H, blk.W = H, W
-            if self.use_checkpoint:
+            if self.grad_checkpointing and not torch.jit.is_scripting():
                 x = checkpoint.checkpoint(blk, x)
             else:
                 x = blk(x)
-
-        if self.downsample is not None:
-            x = x.transpose(1, 2).reshape(x.shape[0], -1, H, W)
-            x, Ho, Wo = self.downsample(x)
-        else:
-            Ho, Wo = H, W
-        return x, Ho, Wo
-
-    def extra_repr(self) -> str:
-        return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}"
+        return x
 
 
-class PatchEmbed(nn.Module):
-    r""" Image to Patch Embedding
-
+class Downsample(nn.Module):
+    r"""
     Args:
-        img_size (int): Image size.  Default: 224.
-        patch_size (int): Patch token size. Default: 4.
-        in_chans (int): Number of input image channels. Default: 3.
-        embed_dim (int): Number of linear projection output channels. Default: 96.
+        in_chs (int): Number of input image channels
+        out_chs (int): Number of linear projection output channels
+        stride (int): Downsample stride. Default: 4.
         norm_layer (nn.Module, optional): Normalization layer. Default: None
     """
 
     def __init__(
             self,
-            img_size=(224, 224),
-            patch_size=4,
-            in_chans=3,
-            embed_dim=96,
-            use_conv_embed=False,
+            in_chs,
+            out_chs,
+            stride=4,
+            overlap=False,
             norm_layer=None,
-            is_stem=False,
     ):
         super().__init__()
-        patch_size = to_2tuple(patch_size)
-        patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]]
-        self.img_size = img_size
-        self.patch_size = patch_size
-        self.patches_resolution = patches_resolution
-        self.num_patches = patches_resolution[0] * patches_resolution[1]
-
-        self.in_chans = in_chans
-        self.embed_dim = embed_dim
-
+        self.stride = stride
         padding = 0
-        kernel_size = patch_size
-        stride = patch_size
-        if use_conv_embed:
-            # if we choose to use conv embedding, then we treat the stem and non-stem differently
-            if is_stem:
-                kernel_size = 7
-                padding = 2
-                stride = 4
-            else:
-                kernel_size = 3
-                padding = 1
-                stride = 2
-        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=kernel_size, stride=stride, padding=padding)
-
-        if norm_layer is not None:
-            self.norm = norm_layer(embed_dim)
-        else:
-            self.norm = None
+        kernel_size = stride
+        if overlap:
+            assert stride in (2, 4)
+            if stride == 4:
+                kernel_size, padding = 7, 2
+            elif stride == 2:
+                kernel_size, padding = 3, 1
+        self.proj = nn.Conv2d(in_chs, out_chs, kernel_size=kernel_size, stride=stride, padding=padding)
+        self.norm = norm_layer(out_chs) if norm_layer is not None else nn.Identity()
 
     def forward(self, x):
-        B, C, H, W = x.shape
         x = self.proj(x)
-        H, W = x.shape[2:]
-        x = x.flatten(2).transpose(1, 2)  # B Ph*Pw C
-        if self.norm is not None:
-            x = self.norm(x)
-        return x, H, W
+        x = self.norm(x)
+        return x
 
 
 class FocalNet(nn.Module):
     r""" Focal Modulation Networks (FocalNets)
 
     Args:
-        img_size (int | tuple(int)): Input image size. Default 224
-        patch_size (int | tuple(int)): Patch size. Default: 4
         in_chans (int): Number of input image channels. Default: 3
         num_classes (int): Number of classes for classification head. Default: 1000
         embed_dim (int): Patch embedding dimension. Default: 96
         depths (tuple(int)): Depth of each Focal Transformer layer.
         mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4
-        drop_rate (float): Dropout rate. Default: 0
-        drop_path_rate (float): Stochastic depth rate. Default: 0.1
-        norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm.
-        patch_norm (bool): If True, add normalization after patch embedding. Default: True
-        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False
         focal_levels (list): How many focal levels at all stages. Note that this excludes the finest-grain level.
             Default: [1, 1, 1, 1]
         focal_windows (list): The focal window size at all stages. Default: [7, 5, 3, 1]
-        use_conv_embed (bool): Whether to use convolutional embedding.
+        use_overlap_down (bool): Whether to use convolutional embedding.
+        use_post_norm (bool): Whether to use layernorm after modulation (it helps stablize training of large models)
         layerscale_value (float): Value for layer scale. Default: 1e-4
-        use_postln (bool): Whether to use layernorm after modulation (it helps stablize training of large models)
+        drop_rate (float): Dropout rate. Default: 0
+        drop_path_rate (float): Stochastic depth rate. Default: 0.1
+        norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm.
+
     """
 
     def __init__(
             self,
-            img_size=224,
-            patch_size=4,
             in_chans=3,
             num_classes=1000,
+            global_pool='avg',
             embed_dim=96,
-            depths=[2, 2, 6, 2],
+            depths=(2, 2, 6, 2),
             mlp_ratio=4.,
-            drop_rate=0.,
-            drop_path_rate=0.1,
-            norm_layer=nn.LayerNorm,
-            patch_norm=True,
-            use_checkpoint=False,
-            focal_levels=[2, 2, 2, 2],
-            focal_windows=[3, 3, 3, 3],
-            use_conv_embed=False,
-            layerscale_value=None,
-            use_postln=False,
-            use_postln_in_modulation=False,
+            focal_levels=(2, 2, 2, 2),
+            focal_windows=(3, 3, 3, 3),
+            use_overlap_down=False,
+            use_post_norm=False,
+            use_post_norm_in_modulation=False,
             normalize_modulator=False,
+            head_hidden_size=None,
+            head_init_scale=1.0,
+            layerscale_value=None,
+            drop_rate=0.,
+            proj_drop_rate=0.,
+            drop_path_rate=0.1,
+            norm_layer=partial(LayerNorm2d, eps=1e-5),
             **kwargs,
     ):
         super().__init__()
@@ -407,129 +363,186 @@ class FocalNet(nn.Module):
 
         self.num_classes = num_classes
         self.embed_dim = embed_dim
-        self.patch_norm = patch_norm
         self.num_features = embed_dim[-1]
-        self.mlp_ratio = mlp_ratio
+        self.feature_info = []
 
-        # split image into patches using either non-overlapped embedding or overlapped embedding
-        self.patch_embed = PatchEmbed(
-            img_size=to_2tuple(img_size),
-            patch_size=patch_size,
-            in_chans=in_chans,
-            embed_dim=embed_dim[0],
-            use_conv_embed=use_conv_embed,
-            norm_layer=norm_layer if self.patch_norm else None,
-            is_stem=True
+        self.stem = Downsample(
+            in_chs=in_chans,
+            out_chs=embed_dim[0],
+            overlap=use_overlap_down,
+            norm_layer=norm_layer,
         )
+        in_dim = embed_dim[0]
 
-        num_patches = self.patch_embed.num_patches
-        patches_resolution = self.patch_embed.patches_resolution
-        self.patches_resolution = patches_resolution
-        self.pos_drop = nn.Dropout(p=drop_rate)
-
-        # stochastic depth
         dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))]  # stochastic depth decay rule
-
-        # build layers
-        self.layers = nn.ModuleList()
+        layers = []
         for i_layer in range(self.num_layers):
+            out_dim = embed_dim[i_layer]
             layer = BasicLayer(
-                dim=embed_dim[i_layer],
-                out_dim=embed_dim[i_layer + 1] if (i_layer < self.num_layers - 1) else None,
-                input_resolution=(
-                    patches_resolution[0] // (2 ** i_layer), patches_resolution[1] // (2 ** i_layer)),
+                dim=in_dim,
+                out_dim=out_dim,
                 depth=depths[i_layer],
-                mlp_ratio=self.mlp_ratio,
-                drop=drop_rate,
-                drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])],
-                norm_layer=norm_layer,
-                downsample=PatchEmbed if (i_layer < self.num_layers - 1) else None,
+                mlp_ratio=mlp_ratio,
+                downsample=i_layer > 0,
                 focal_level=focal_levels[i_layer],
                 focal_window=focal_windows[i_layer],
-                use_conv_embed=use_conv_embed,
-                use_checkpoint=use_checkpoint,
+                use_overlap_down=use_overlap_down,
+                use_post_norm=use_post_norm,
+                use_post_norm_in_modulation=use_post_norm_in_modulation,
+                normalize_modulator=normalize_modulator,
                 layerscale_value=layerscale_value,
-                use_postln=use_postln,
-                use_postln_in_modulation=use_postln_in_modulation,
-                normalize_modulator=normalize_modulator
-                )
-            self.layers.append(layer)
+                proj_drop=proj_drop_rate,
+                drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])],
+                norm_layer=norm_layer,
+            )
+            in_dim = out_dim
+            layers += [layer]
+            self.feature_info += [dict(num_chs=out_dim, reduction=4 * 2 ** i_layer, module=f'layers.{i_layer}')]
 
-        self.norm = norm_layer(self.num_features)
-        self.avgpool = nn.AdaptiveAvgPool1d(1)
-        self.head = nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity()
+        self.layers = nn.Sequential(*layers)
 
-        self.apply(self._init_weights)
+        if head_hidden_size:
+            self.norm = nn.Identity()
+            self.head = NormMlpClassifierHead(
+                self.num_features,
+                num_classes,
+                hidden_size=head_hidden_size,
+                pool_type=global_pool,
+                drop_rate=drop_rate,
+                norm_layer=norm_layer,
+            )
+        else:
+            self.norm = norm_layer(self.num_features)
+            self.head = ClassifierHead(
+                self.num_features,
+                num_classes,
+                pool_type=global_pool,
+                drop_rate=drop_rate
+            )
 
-    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)
+        named_apply(partial(_init_weights, head_init_scale=head_init_scale), self)
 
     @torch.jit.ignore
     def no_weight_decay(self):
         return {''}
 
-    def forward_features(self, x):
-        x, H, W = self.patch_embed(x)
-        x = self.pos_drop(x)
+    @torch.jit.ignore
+    def set_grad_checkpointing(self, enable=True):
+        self.grad_checkpointing = enable
+        for l in self.layers:
+            l.set_grad_checkpointing(enable=enable)
 
-        for layer in self.layers:
-            x, H, W = layer(x, H, W)
-        x = self.norm(x)  # B L C
-        x = self.avgpool(x.transpose(1, 2))  # B C 1
-        x = torch.flatten(x, 1)
+    @torch.jit.ignore
+    def get_classifier(self):
+        return self.classifier.fc
+
+    def reset_classifier(self, num_classes, global_pool=None):
+        self.classifier.reset(num_classes, global_pool=global_pool)
+
+    def forward_features(self, x):
+        x = self.stem(x)
+        x = self.layers(x)
+        x = self.norm(x)
         return x
 
+    def forward_head(self, x, pre_logits: bool = False):
+        return self.head(x, pre_logits=pre_logits)
+
     def forward(self, x):
         x = self.forward_features(x)
-        x = self.head(x)
+        x = self.forward_head(x)
         return x
 
 
+def _init_weights(module, name=None, head_init_scale=1.0):
+    if isinstance(module, nn.Conv2d):
+        trunc_normal_(module.weight, std=.02)
+        if module.bias is not None:
+            nn.init.zeros_(module.bias)
+    elif isinstance(module, nn.Linear):
+        trunc_normal_(module.weight, std=.02)
+        if module.bias is not None:
+            nn.init.zeros_(module.bias)
+        if name and 'head.fc' in name:
+            module.weight.data.mul_(head_init_scale)
+            module.bias.data.mul_(head_init_scale)
+
+
 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,
+        'crop_pct': .9, 'interpolation': 'bicubic',
         'mean': IMAGENET_DEFAULT_MEAN, 'std': IMAGENET_DEFAULT_STD,
-        'first_conv': 'patch_embed.proj', 'classifier': 'head',
+        'first_conv': 'stem.proj', 'classifier': 'head.fc',
         **kwargs
     }
 
 
 default_cfgs = {
-    "focalnet_tiny_srf": _cfg(),
-    "focalnet_small_srf": _cfg(url="https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_small_srf.pth"),
-    "focalnet_base_srf": _cfg(),
-    "focalnet_tiny_lrf": _cfg(),
-    "focalnet_small_lrf": _cfg(),
-    "focalnet_base_lrf": _cfg(url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_base_lrf.pth'),
-    "focalnet_large_fl3": _cfg(url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_large_lrf_384.pth', input_size=(3, 384, 384), num_classes=21842),
-    "focalnet_large_fl4": _cfg(url="https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_large_lrf_384_fl4.pth", input_size=(3, 384, 384), num_classes=21842),
+    "focalnet_tiny_srf": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_tiny_srf.pth'),
+    "focalnet_small_srf": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_small_srf.pth'),
+    "focalnet_base_srf": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_base_srf.pth'),
+    "focalnet_tiny_lrf": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_tiny_lrf.pth'),
+    "focalnet_small_lrf": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_small_lrf.pth'),
+    "focalnet_base_lrf": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_base_lrf.pth'),
+    "focalnet_large_fl3": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_large_lrf_384.pth',
+        input_size=(3, 384, 384), crop_pct=1.0, num_classes=21842),
+    "focalnet_large_fl4": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_large_lrf_384_fl4.pth',
+        input_size=(3, 384, 384), crop_pct=1.0, num_classes=21842),
+    "focalnet_xlarge_fl3": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_xlarge_lrf_384.pth',
+        input_size=(3, 384, 384), crop_pct=1.0, num_classes=21842),
+    "focalnet_xlarge_fl4": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_xlarge_lrf_384_fl4.pth',
+        input_size=(3, 384, 384), crop_pct=1.0, num_classes=21842),
+    "focalnet_huge_fl3": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_huge_lrf_224.pth',
+        num_classes=0),
+    "focalnet_huge_fl4": _cfg(
+        url='https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_huge_lrf_224_fl4.pth',
+        num_classes=0),
 }
 
 
-def checkpoint_filter_fn(state_dict, model):
+def checkpoint_filter_fn(state_dict, model: FocalNet):
+    if 'stem.proj.weight' in state_dict:
+        return
+    import re
     out_dict = {}
     if 'model' in state_dict:
-        # For deit models
         state_dict = state_dict['model']
+    dest_dict = model.state_dict()
     for k, v in state_dict.items():
-        if any([n in k for n in ('relative_position_index', 'relative_coords_table')]):
-            continue  # skip buffers that should not be persistent
+        k = re.sub(r'gamma_([0-9])', r'ls\1.gamma', k)
+        k = k.replace('patch_embed', 'stem')
+        k = re.sub(r'layers.(\d+).downsample', lambda x: f'layers.{int(x.group(1)) + 1}.downsample', k)
+        if 'norm' in k and k not in dest_dict:
+            k = re.sub(r'norm([0-9])', r'norm\1_post', k)
+        k = k.replace('ln.', 'norm.')
+        k = k.replace('head', 'head.fc')
+        if dest_dict[k].shape != v.shape:
+            v = v.reshape(dest_dict[k].shape)
         out_dict[k] = v
     return out_dict
 
 
 def _create_focalnet(variant, pretrained=False, **kwargs):
+    default_out_indices = tuple(i for i, _ in enumerate(kwargs.get('depths', (1, 1, 3, 1))))
+    out_indices = kwargs.pop('out_indices', default_out_indices)
+
     model = build_model_with_cfg(
         FocalNet, variant, pretrained,
         pretrained_filter_fn=checkpoint_filter_fn,
+        feature_cfg=dict(flatten_sequential=True, out_indices=out_indices),
         **kwargs)
     return model
 
@@ -569,10 +582,13 @@ def focalnet_base_lrf(pretrained=False, **kwargs):
     model_kwargs = dict(depths=[2, 2, 18, 2], embed_dim=128, focal_levels=[3, 3, 3, 3], **kwargs)
     return _create_focalnet('focalnet_base_lrf', pretrained=pretrained, **model_kwargs)
 
+
 # FocalNet large+ models
 @register_model
 def focalnet_large_fl3(pretrained=False, **kwargs):
-    model_kwargs = dict(depths=[2, 2, 18, 2], embed_dim=192, focal_levels=[3, 3, 3, 3], **kwargs)
+    model_kwargs = dict(
+        depths=[2, 2, 18, 2], embed_dim=192, focal_levels=[3, 3, 3, 3], focal_windows=[5] * 4,
+        use_post_norm=True, use_overlap_down=True, layerscale_value=1e-4, **kwargs)
     return _create_focalnet('focalnet_large_fl3', pretrained=pretrained, **model_kwargs)
 
 
@@ -580,37 +596,38 @@ def focalnet_large_fl3(pretrained=False, **kwargs):
 def focalnet_large_fl4(pretrained=False, **kwargs):
     model_kwargs = dict(
         depths=[2, 2, 18, 2], embed_dim=192, focal_levels=[4, 4, 4, 4],
-        use_conv_embed=True, layerscale_value=1e-4, **kwargs)
+        use_post_norm=True, use_overlap_down=True, layerscale_value=1e-4, **kwargs)
     return _create_focalnet('focalnet_large_fl4', pretrained=pretrained, **model_kwargs)
 
-#
-# @register_model
-# def focalnet_large_fl4(pretrained=False, **kwargs):
-#     model_kwargs = dict(depths=[2, 2, 18, 2], embed_dim=192, focal_levels=[4, 4, 4, 4], **kwargs)
-#     return _create_focalnet('focalnet_large_fl4', pretrained=pretrained, **model_kwargs)
-
-
 
 @register_model
 def focalnet_xlarge_fl3(pretrained=False, **kwargs):
-    model_kwargs = dict(depths=[2, 2, 18, 2], embed_dim=256, focal_levels=[3, 3, 3, 3], **kwargs)
+    model_kwargs = dict(
+        depths=[2, 2, 18, 2], embed_dim=256, focal_levels=[3, 3, 3, 3], focal_windows=[5] * 4,
+        use_post_norm=True, use_overlap_down=True, layerscale_value=1e-4, **kwargs)
     return _create_focalnet('focalnet_xlarge_fl3', pretrained=pretrained, **model_kwargs)
 
 
 @register_model
 def focalnet_xlarge_fl4(pretrained=False, **kwargs):
-    model_kwargs = dict(depths=[2, 2, 18, 2], embed_dim=256, focal_levels=[4, 4, 4, 4], **kwargs)
+    model_kwargs = dict(
+        depths=[2, 2, 18, 2], embed_dim=256, focal_levels=[4, 4, 4, 4],
+        use_post_norm=True, use_overlap_down=True, layerscale_value=1e-4, **kwargs)
     return _create_focalnet('focalnet_xlarge_fl4', pretrained=pretrained, **model_kwargs)
 
 
 @register_model
 def focalnet_huge_fl3(pretrained=False, **kwargs):
-    model_kwargs = dict(depths=[2, 2, 18, 2], embed_dim=352, focal_levels=[3, 3, 3, 3], **kwargs)
+    model_kwargs = dict(
+        depths=[2, 2, 18, 2], embed_dim=352, focal_levels=[3, 3, 3, 3], focal_windows=[3] * 4,
+        use_post_norm=True, use_post_norm_in_modulation=True, use_overlap_down=True, layerscale_value=1e-4, **kwargs)
     return _create_focalnet('focalnet_huge_fl3', pretrained=pretrained, **model_kwargs)
 
 
 @register_model
 def focalnet_huge_fl4(pretrained=False, **kwargs):
-    model_kwargs = dict(depths=[2, 2, 18, 2], embed_dim=352, focal_levels=[4, 4, 4, 4], **kwargs)
+    model_kwargs = dict(
+        depths=[2, 2, 18, 2], embed_dim=352, focal_levels=[4, 4, 4, 4],
+        use_post_norm=True, use_post_norm_in_modulation=True, use_overlap_down=True, layerscale_value=1e-4, **kwargs)
     return _create_focalnet('focalnet_huge_fl4', pretrained=pretrained, **model_kwargs)