From 1b7671439741f35d00251140c5a021af781e5ca1 Mon Sep 17 00:00:00 2001
From: Fredo Guan <fredo.guan@hotmail.com>
Date: Sat, 10 Dec 2022 04:16:11 -0800
Subject: [PATCH] Update davit.py

---
 timm/models/davit.py | 39 +++++++++++++++++++++++++++++++--------
 1 file changed, 31 insertions(+), 8 deletions(-)

diff --git a/timm/models/davit.py b/timm/models/davit.py
index 076e84b5..f019c5d8 100644
--- a/timm/models/davit.py
+++ b/timm/models/davit.py
@@ -514,7 +514,7 @@ class DaViT(nn.Module):
             self.feature_info += [dict(num_chs=self.embed_dims[stage_id], reduction=2, module=f'stages.{stage_id}')]
         
         
-        self.stages = nn.ModuleList(stages)
+        self.stages = SequentialWithSize(*stages)
         
         self.norms = norm_layer(self.num_features)
         self.head = ClassifierHead(self.num_features, num_classes, pool_type=global_pool, drop_rate=drop_rate)
@@ -544,7 +544,7 @@ class DaViT(nn.Module):
             global_pool = self.head.global_pool.pool_type
         self.head = ClassifierHead(self.num_features, num_classes, pool_type=global_pool, drop_rate=self.drop_rate)
     
-    
+    '''
     def forward_network(self, x : Tensor):
         size: Tuple[int, int] = (x.size(2), x.size(3))
         features = [x]
@@ -570,12 +570,19 @@ class DaViT(nn.Module):
             outs.append(out.view(-1, H, W, self.embed_dims[i]).permute(0, 3, 1, 2).contiguous())
         
         return outs
-
+    '''
+    
+    
+    
+    
     def forward_features(self, x):
-        x, sizes = self.forward_network(x)
+        #x, sizes = self.forward_network(x)
+        size: Tuple[int, int] = (x.size(2), x.size(3))
+        x, size = stages(x, size)
+        
         # take final feature and norm
-        x = self.norms(x[-1])
-        H, W = sizes[-1]
+        x = self.norms(x)
+        H, W = sizes
         x = x.view(-1, H, W, self.embed_dims[-1]).permute(0, 3, 1, 2).contiguous()
         return x
     
@@ -590,7 +597,8 @@ class DaViT(nn.Module):
     def forward(self, x):
         return self.forward_classifier(x)
         
-        
+      
+'''      
 class DaViTFeatures(DaViT):
 
     def __init__(self, *args, **kwargs):
@@ -600,6 +608,8 @@ class DaViTFeatures(DaViT):
     def forward(self, x) -> List[Tensor]:
         return self.forward_pyramid_features(x)
     
+'''
+
 
 def checkpoint_filter_fn(state_dict, model):
     """ Remap MSFT checkpoints -> timm """
@@ -620,7 +630,7 @@ def checkpoint_filter_fn(state_dict, model):
     return out_dict
     
     
-    
+'''    
 def _create_davit(variant, pretrained=False, **kwargs):
     model_cls = DaViT
     features_only = False
@@ -643,6 +653,19 @@ def _create_davit(variant, pretrained=False, **kwargs):
         model.default_cfg = model.pretrained_cfg  # backwards compat
     return model
 
+'''
+def _create_davit(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(
+        DaViT,
+        variant,
+        pretrained,
+        pretrained_filter_fn=checkpoint_filter_fn,
+        feature_cfg=dict(flatten_sequential=True, out_indices=out_indices),
+        **kwargs)
+
+    return model
 
 
 def _cfg(url='', **kwargs): # not sure how this should be set up