Fix timm incompatibility

Backports isl-org/MiDaS#234
Fixes #319
Fixes #323

dmidas/backbones/beit.py

@@ -44,7 +44,7 @@ def _get_rel_pos_bias(self, window_size):
     old_sub_table = old_relative_position_bias_table[:old_num_relative_distance - 3]
 
     old_sub_table = old_sub_table.reshape(1, old_width, old_height, -1).permute(0, 3, 1, 2)
-    new_sub_table = F.interpolate(old_sub_table, size=(new_height, new_width), mode="bilinear")
+    new_sub_table = F.interpolate(old_sub_table, size=(int(new_height),int(new_width)), mode="bilinear")
     new_sub_table = new_sub_table.permute(0, 2, 3, 1).reshape(new_num_relative_distance - 3, -1)
 
     new_relative_position_bias_table = torch.cat(
@@ -96,12 +96,12 @@ def block_forward(self, x, resolution, shared_rel_pos_bias: Optional[torch.Tenso
     Modification of timm.models.beit.py: Block.forward to support arbitrary window sizes.
     """
     if self.gamma_1 is None:
-        x = x + self.drop_path(self.attn(self.norm1(x), resolution, shared_rel_pos_bias=shared_rel_pos_bias))
-        x = x + self.drop_path(self.mlp(self.norm2(x)))
+        x = x + self.drop_path1(self.attn(self.norm1(x), resolution, shared_rel_pos_bias=shared_rel_pos_bias))
+        x = x + self.drop_path2(self.mlp(self.norm2(x)))
     else:
-        x = x + self.drop_path(self.gamma_1 * self.attn(self.norm1(x), resolution,
+        x = x + self.drop_path1(self.gamma_1 * self.attn(self.norm1(x), resolution,
                                                         shared_rel_pos_bias=shared_rel_pos_bias))
-        x = x + self.drop_path(self.gamma_2 * self.mlp(self.norm2(x)))
+        x = x + self.drop_path2(self.gamma_2 * self.mlp(self.norm2(x)))
     return x
 
 

dzoedepth/models/base_models/midas.py

@@ -338,7 +338,9 @@ def build(midas_model_type="DPT_BEiT_L_384", train_midas=False, use_pretrained_m
             kwargs = MidasCore.parse_img_size(kwargs)
         img_size = kwargs.pop("img_size", [384, 384])
         print("img_size", img_size)
-        midas = torch.hub.load("intel-isl/MiDaS", midas_model_type,
+        # TODO: use locally-bundled midas
+        # The repo should be changed back to isl-org/MiDaS once this MR lands
+        midas = torch.hub.load("AyaanShah2204/MiDaS", midas_model_type,
                                pretrained=use_pretrained_midas, force_reload=force_reload)
         kwargs.update({'keep_aspect_ratio': force_keep_ar})
         midas_core = MidasCore(midas, trainable=train_midas, fetch_features=fetch_features,

install.py

@@ -22,8 +22,8 @@ def ensure(module_name, min_version=None):
     msg = f'{requirement} requirement for depthmap script'
     launch.run_pip(cmd, msg)
 
-if not launch.is_installed("timm"): #0.6.7  #  For midas
-    launch.run_pip('install --force-reinstall "timm==0.6.12"', "timm requirement for depthmap script")
+
+ensure('timm', '0.9.2')  # Just in case
 
 ensure('matplotlib')
 
@@ -47,8 +47,8 @@ def ensure(module_name, min_version=None):
 
 if not launch.is_installed("networkx"):
     launch.run_pip('install install "networkx==2.5"', "networkx requirement for depthmap script")
+
 if platform.system() == 'Windows':
     ensure('pyqt5')
-
 if platform.system() == 'Darwin':
     ensure('pyqt6')