update riformer mmpretrain

configs/riformer/riformer-m48_32xb128_in1k.py

@@ -27,7 +27,6 @@ model = dict(
         loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
     ))
 
-
 # schedule settings
 optim_wrapper = dict(
     optimizer=dict(lr=4e-3),

configs/riformer/riformer-m48_32xb128_in1k_384.py

@@ -27,7 +27,6 @@ model = dict(
         loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
     ))
 
-
 # schedule settings
 optim_wrapper = dict(
     optimizer=dict(lr=4e-3),

configs/riformer/riformer-s12_32xb128_in1k.py

@@ -27,7 +27,6 @@ model = dict(
         loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
     ))
 
-
 # schedule settings
 optim_wrapper = dict(
     optimizer=dict(lr=4e-3),

configs/riformer/riformer-s12_32xb128_in1k_384.py

@@ -27,7 +27,6 @@ model = dict(
         loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
     ))
 
-
 # schedule settings
 optim_wrapper = dict(
     optimizer=dict(lr=4e-3),

mmpretrain/models/backbones/riformer.py

@@ -3,67 +3,12 @@ from typing import Sequence
 
 import torch
 import torch.nn as nn
-from mmcv.cnn.bricks import DropPath, build_activation_layer, build_norm_layer
+from mmcv.cnn.bricks import DropPath, build_norm_layer
 from mmengine.model import BaseModule
 
 from mmpretrain.registry import MODELS
 from .base_backbone import BaseBackbone
-
-
-class PatchEmbed(nn.Module):
-    """Patch Embedding module implemented by a layer of convolution.
-
-    Input: tensor in shape [B, C, H, W]
-    Output: tensor in shape [B, C, H/stride, W/stride]
-    Args:
-        patch_size (int): Patch size of the patch embedding. Defaults to 16.
-        stride (int): Stride of the patch embedding. Defaults to 16.
-        padding (int): Padding of the patch embedding. Defaults to 0.
-        in_chans (int): Input channels. Defaults to 3.
-        embed_dim (int): Output dimension of the patch embedding.
-            Defaults to 768.
-        norm_layer (module): Normalization module. Defaults to None (not use).
-    """
-
-    def __init__(self,
-                 patch_size=16,
-                 stride=16,
-                 padding=0,
-                 in_chans=3,
-                 embed_dim=768,
-                 norm_layer=None):
-        super().__init__()
-        self.proj = nn.Conv2d(
-            in_chans,
-            embed_dim,
-            kernel_size=patch_size,
-            stride=stride,
-            padding=padding)
-        self.norm = norm_layer(embed_dim) if norm_layer else nn.Identity()
-
-    def forward(self, x):
-        x = self.proj(x)
-        x = self.norm(x)
-        return x
-
-
-class Pooling(nn.Module):
-    """Pooling module.
-
-    Args:
-        pool_size (int): Pooling size. Defaults to 3.
-    """
-
-    def __init__(self, pool_size=3):
-        super().__init__()
-        self.pool = nn.AvgPool2d(
-            pool_size,
-            stride=1,
-            padding=pool_size // 2,
-            count_include_pad=False)
-
-    def forward(self, x):
-        return self.pool(x) - x
+from .poolformer import Mlp, PatchEmbed
 
 
 class Affine(nn.Module):
@@ -88,43 +33,6 @@ class Affine(nn.Module):
         return self.affine(x) - x
 
 
-class Mlp(nn.Module):
-    """Mlp implemented by with 1*1 convolutions.
-
-    Input: Tensor with shape [B, C, H, W].
-    Output: Tensor with shape [B, C, H, W].
-    Args:
-        in_features (int): Dimension of input features.
-        hidden_features (int): Dimension of hidden features.
-        out_features (int): Dimension of output features.
-        act_cfg (dict): The config dict for activation between pointwise
-            convolution. Defaults to ``dict(type='GELU')``.
-        drop (float): Dropout rate. Defaults to 0.0.
-    """
-
-    def __init__(self,
-                 in_features,
-                 hidden_features=None,
-                 out_features=None,
-                 act_cfg=dict(type='GELU'),
-                 drop=0.):
-        super().__init__()
-        out_features = out_features or in_features
-        hidden_features = hidden_features or in_features
-        self.fc1 = nn.Conv2d(in_features, hidden_features, 1)
-        self.act = build_activation_layer(act_cfg)
-        self.fc2 = nn.Conv2d(hidden_features, out_features, 1)
-        self.drop = nn.Dropout(drop)
-
-    def forward(self, x):
-        x = self.fc1(x)
-        x = self.act(x)
-        x = self.drop(x)
-        x = self.fc2(x)
-        x = self.drop(x)
-        return x
-
-
 class RIFormerBlock(BaseModule):
     """RIFormer Block.
 

tools/model_converters/reparameterize_model.py

@@ -3,6 +3,7 @@ import argparse
 from pathlib import Path
 
 import torch
+
 from mmpretrain.apis import init_model
 from mmpretrain.models.classifiers import ImageClassifier