Use transformers init everywhere (#49573)

Author: guarin

src/transformers/models/gemma4_unified/modeling_gemma4_unified.py

@@ -23,8 +23,8 @@
 
 import torch
 from torch import nn
-from torch.nn import init
 
+from ... import initialization as init
 from ...activations import ACT2FN
 from ...cache_utils import Cache, DynamicCache
 from ...configuration_utils import PreTrainedConfig

src/transformers/models/gemma4_unified/modular_gemma4_unified.py

@@ -19,9 +19,9 @@
 import torch
 from huggingface_hub.dataclasses import strict
 from torch import nn
-from torch.nn import init
 from torchvision.transforms.v2 import functional as tvF
 
+from ... import initialization as init
 from ...cache_utils import Cache, DynamicCache
 from ...configuration_utils import PreTrainedConfig
 from ...image_processing_utils import BatchFeature

src/transformers/models/granite_swa/modeling_granite_swa.py

@@ -24,8 +24,8 @@
 import torch
 from torch import nn
 from torch.nn import functional as F
-from torch.nn import init
 
+from ... import initialization as init
 from ...activations import ACT2FN
 from ...cache_utils import Cache, DynamicCache
 from ...generation import GenerationMixin

src/transformers/models/granite_swa/modular_granite_swa.py

@@ -31,8 +31,8 @@
 from huggingface_hub.dataclasses import strict
 from torch import nn
 from torch.nn import functional as F
-from torch.nn import init
 
+from ... import initialization as init
 from ...cache_utils import Cache, DynamicCache
 from ...masking_utils import create_causal_mask, create_sliding_window_causal_mask
 from ...modeling_outputs import BaseModelOutputWithPast

src/transformers/models/laguna/modeling_laguna.py

@@ -482,7 +482,7 @@ def _init_weights(self, module):
         elif isinstance(module, LagunaTopKRouter):
             init.normal_(module.weight, mean=0.0, std=std)
         if isinstance(module, LagunaTopKRouter):
-            torch.nn.init.zeros_(module.e_score_correction_bias)
+            init.zeros_(module.e_score_correction_bias)
         elif isinstance(module, LagunaRotaryEmbedding):
             for layer_type in module.layer_types:
                 rope_init_fn = module.compute_default_rope_parameters

src/transformers/models/laguna/modular_laguna.py

@@ -362,7 +362,7 @@ class LagunaPreTrainedModel(Qwen2MoePreTrainedModel):
     def _init_weights(self, module):
         super()._init_weights(module)
         if isinstance(module, LagunaTopKRouter):
-            torch.nn.init.zeros_(module.e_score_correction_bias)
+            init.zeros_(module.e_score_correction_bias)
         elif isinstance(module, LagunaRotaryEmbedding):
             for layer_type in module.layer_types:
                 rope_init_fn = module.compute_default_rope_parameters

src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py

@@ -25,6 +25,7 @@
 import torch.nn as nn
 from torch import Tensor
 
+from ... import initialization as init
 from ...activations import ACT2FN
 from ...cache_utils import Cache, DynamicCache, EncoderDecoderCache
 from ...generation import GenerationMixin
@@ -339,7 +340,7 @@ def _get_feat_extract_output_lengths(self, input_lengths: torch.LongTensor) -> t
 
     def _init_weights(self, module: nn.Module):
         if isinstance(module, MoonshineStreamingLayerNorm):
-            nn.init.constant_(module.gamma, 1.0 - module.unit_offset)
+            init.constant_(module.gamma, 1.0 - module.unit_offset)
         else:
             super()._init_weights(module)
 

src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py

@@ -19,6 +19,7 @@
 import torch.nn as nn
 from torch import Tensor
 
+from ... import initialization as init
 from ...cache_utils import Cache
 from ...masking_utils import create_bidirectional_mask
 from ...modeling_outputs import (
@@ -254,7 +255,7 @@ def _get_feat_extract_output_lengths(self, input_lengths: torch.LongTensor) -> t
 
     def _init_weights(self, module: nn.Module):
         if isinstance(module, MoonshineStreamingLayerNorm):
-            nn.init.constant_(module.gamma, 1.0 - module.unit_offset)
+            init.constant_(module.gamma, 1.0 - module.unit_offset)
         else:
             super()._init_weights(module)
 

src/transformers/models/pp_lcnet_v3/modeling_pp_lcnet_v3.py

@@ -23,6 +23,7 @@
 import torch.nn as nn
 from torch import Tensor
 
+from ... import initialization as init
 from ...activations import ACT2FN
 from ...backbone_utils import BackboneMixin, filter_output_hidden_states
 from ...modeling_layers import GradientCheckpointingLayer
@@ -314,8 +315,8 @@ class PPLCNetV3PreTrainedModel(PreTrainedModel):
     def _init_weights(self, module):
         super()._init_weights(module)
         if isinstance(module, (PPLCNetV3LearnableAffineBlock)):
-            nn.init.ones_(module.scale)
-            nn.init.zeros_(module.bias)
+            init.ones_(module.scale)
+            init.zeros_(module.bias)
 
 
 @auto_docstring(

src/transformers/models/pp_lcnet_v3/modular_pp_lcnet_v3.py

@@ -17,6 +17,7 @@
 import torch.nn as nn
 from huggingface_hub.dataclasses import strict
 
+from ... import initialization as init
 from ...activations import ACT2FN
 from ...configuration_utils import PreTrainedConfig
 from ...modeling_utils import PreTrainedModel
@@ -237,8 +238,8 @@ class PPLCNetV3PreTrainedModel(PPLCNetPreTrainedModel):
     def _init_weights(self, module):
         PreTrainedModel._init_weights(module)
         if isinstance(module, (PPLCNetV3LearnableAffineBlock)):
-            nn.init.ones_(module.scale)
-            nn.init.zeros_(module.bias)
+            init.ones_(module.scale)
+            init.zeros_(module.bias)
 
 
 class PPLCNetV3Backbone(PPLCNetBackbone):