fix import

Author: Cyrilvallez

src/transformers/models/bit/modeling_bit.py

@@ -637,7 +637,7 @@ def _init_weights(self, module):
         elif isinstance(module, nn.Linear):
             init.kaiming_uniform_(module.weight, a=math.sqrt(5))
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, (nn.BatchNorm2d, nn.GroupNorm)):

src/transformers/models/deprecated/trajectory_transformer/modeling_trajectory_transformer.py

@@ -98,7 +98,7 @@ def _init_weights(self, module):
             for i in range(module.n_models):
                 init.kaiming_uniform_(module.weight[i], a=math.sqrt(5) / self.config.kaiming_initializer_range)
                 if module.bias is not None:
-                    fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight[i])
+                    fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight[i])
                     bound = (1 / math.sqrt(fan_in)) * self.config.initializer_range
                     init.uniform_(module.bias[i], -bound, bound)
 
@@ -161,7 +161,7 @@ def reset_parameters(self):
         for i in range(self.n_models):
             init.kaiming_uniform_(self.weight[i], a=math.sqrt(5))
             if self.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(self.weight[i])
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(self.weight[i])
                 bound = 1 / math.sqrt(fan_in)
                 init.uniform_(self.bias[i], -bound, bound)
 

src/transformers/models/emu3/modeling_emu3.py

@@ -944,13 +944,13 @@ def _init_weights(self, module):
         if isinstance(module, (nn.Conv2d, nn.Conv3d)):
             init.kaiming_normal_(module.weight, mode="fan_out", nonlinearity="relu")
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in)
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, nn.Linear):
             init.kaiming_uniform_(module.weight, a=math.sqrt(5))
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, (nn.BatchNorm2d, nn.BatchNorm3d, nn.GroupNorm)):

src/transformers/models/emu3/modular_emu3.py

@@ -694,13 +694,13 @@ def _init_weights(self, module):
         if isinstance(module, (nn.Conv2d, nn.Conv3d)):
             init.kaiming_normal_(module.weight, mode="fan_out", nonlinearity="relu")
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in)
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, nn.Linear):
             init.kaiming_uniform_(module.weight, a=math.sqrt(5))
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, (nn.BatchNorm2d, nn.BatchNorm3d, nn.GroupNorm)):

src/transformers/models/eomt/modeling_eomt.py

@@ -1003,7 +1003,7 @@ def _init_weights(self, module: nn.Module) -> None:
         if isinstance(module, (nn.Linear, nn.Conv2d, nn.ConvTranspose2d)):
             init.kaiming_uniform_(module.weight, a=math.sqrt(5))
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, nn.LayerNorm):

src/transformers/models/eomt/modular_eomt.py

@@ -408,7 +408,7 @@ def _init_weights(self, module: nn.Module) -> None:
         if isinstance(module, (nn.Linear, nn.Conv2d, nn.ConvTranspose2d)):
             init.kaiming_uniform_(module.weight, a=math.sqrt(5))
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, nn.LayerNorm):

src/transformers/models/mamba2/modeling_mamba2.py

@@ -737,7 +737,7 @@ def _init_weights(self, module):
 
             # # Inverse of softplus: https://github.com/pytorch/pytorch/issues/72759
             inv_dt = dt + torch.log(-torch.expm1(-dt))
-            init.copy_(module.dt_proj.bias, inv_dt)
+            init.copy_(module.dt_bias, inv_dt)
 
             init.kaiming_uniform_(module.conv1d.weight, a=math.sqrt(5))
             if module.conv1d.bias is not None:

src/transformers/models/regnet/modeling_regnet.py

@@ -272,7 +272,7 @@ def _init_weights(self, module):
         elif isinstance(module, nn.Linear):
             init.kaiming_uniform_(module.weight, a=math.sqrt(5))
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, (nn.BatchNorm2d, nn.GroupNorm)):

src/transformers/models/resnet/modeling_resnet.py

@@ -259,7 +259,7 @@ def _init_weights(self, module):
         elif isinstance(module, nn.Linear):
             init.kaiming_uniform_(module.weight, a=math.sqrt(5))
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, (nn.BatchNorm2d, nn.GroupNorm)):

src/transformers/models/rt_detr/modeling_rt_detr_resnet.py

@@ -313,7 +313,7 @@ def _init_weights(self, module):
         elif isinstance(module, nn.Linear):
             init.kaiming_uniform_(module.weight, a=math.sqrt(5))
             if module.bias is not None:
-                fan_in, _ = init._calculate_fan_in_and_fan_out(module.weight)
+                fan_in, _ = torch.nn.init._calculate_fan_in_and_fan_out(module.weight)
                 bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
                 init.uniform_(module.bias, -bound, bound)
         elif isinstance(module, (nn.BatchNorm2d, nn.GroupNorm)):