Add more missing arguments (#40354)

Add missing arguments

Signed-off-by: cyy <cyyever@outlook.com>

Author: cyyever

src/transformers/models/bamba/modular_bamba.py

@@ -708,7 +708,7 @@ class BambaRMSNorm(LlamaRMSNorm):
 
 class BambaDecoderLayer(JambaAttentionDecoderLayer):
     def __init__(self, config: BambaConfig, layer_idx: int, layer_type: str = "mamba"):
-        super().__init__()
+        super().__init__(config, layer_idx)
 
         del self.self_attn
 

src/transformers/models/cohere2/modular_cohere2.py

@@ -271,7 +271,7 @@ class Cohere2LayerNorm(CohereLayerNorm):
     pass
 
 
-class Cohere2Attention(CohereAttention, nn.Module):
+class Cohere2Attention(CohereAttention):
     """Multi-headed attention from 'Attention Is All You Need' paper"""
 
     def __init__(self, config: Cohere2Config, layer_idx: Optional[int] = None):

src/transformers/models/d_fine/modular_d_fine.py

@@ -898,7 +898,7 @@ def __init__(self, config: DFineConfig):
 
 class DFineForObjectDetection(RTDetrForObjectDetection, DFinePreTrainedModel):
     def __init__(self, config: DFineConfig):
-        DFinePreTrainedModel.__init__(config)
+        DFinePreTrainedModel.__init__(self, config)
 
         # D-FINE encoder-decoder model
         self.eval_idx = config.eval_idx if config.eval_idx >= 0 else config.decoder_layers + config.eval_idx

src/transformers/models/data2vec/modular_data2vec_audio.py

@@ -112,7 +112,7 @@ def forward(self, hidden_states):
         return hidden_states
 
 
-class Data2VecAudioFeatureEncoder(Wav2Vec2FeatureEncoder, nn.Module):
+class Data2VecAudioFeatureEncoder(Wav2Vec2FeatureEncoder):
     def __init__(self, config):
         nn.Module.__init__(self)
         self.conv_layers = nn.ModuleList(
@@ -183,7 +183,7 @@ def load_adapter(self):
 
 class Data2VecAudioModel(Data2VecAudioPreTrainedModel, Wav2Vec2Model):
     def __init__(self, config: Data2VecAudioConfig):
-        Data2VecAudioPreTrainedModel.__init__(config)
+        Data2VecAudioPreTrainedModel.__init__(self, config)
         self.config = config
         self.feature_extractor = Data2VecAudioFeatureEncoder(config)
         self.feature_projection = Data2VecAudioFeatureProjection(config)
@@ -215,7 +215,7 @@ def forward(self, **super_kwargs):
 
 class Data2VecAudioForCTC(Data2VecAudioPreTrainedModel, Wav2Vec2ForCTC):
     def __init__(self, config):
-        Data2VecAudioPreTrainedModel.__init__(config)
+        Data2VecAudioPreTrainedModel.__init__(self, config)
 
         self.data2vec_audio = Data2VecAudioModel(config)
         self.dropout = nn.Dropout(config.final_dropout)

src/transformers/models/dia/modular_dia.py

@@ -107,7 +107,7 @@ class DiaRotaryEmbedding(LlamaRotaryEmbedding):
     pass
 
 
-class DiaSelfAttention(LlamaAttention, nn.Module):
+class DiaSelfAttention(LlamaAttention):
     """Multi-headed attention from 'Attention Is All You Need' paper"""
 
     def __init__(self, config: Union[DiaEncoderConfig, DiaDecoderConfig], layer_idx: int, is_causal: bool = False):

src/transformers/models/dots1/modular_dots1.py

@@ -62,7 +62,7 @@ class Dots1TopkRouter(DeepseekV3TopkRouter):
 
 class Dots1DecoderLayer(DeepseekV3DecoderLayer):
     def __init__(self, config: Dots1Config, layer_idx: int):
-        super().__init__()
+        super().__init__(config, layer_idx)
         self.attention_type = config.layer_types[layer_idx]
 
 

src/transformers/models/ernie4_5/modular_ernie4_5.py

@@ -84,7 +84,7 @@ def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
 
 class Ernie4_5MLP(LlamaMLP):
     def __init__(self, config: Ernie4_5Config):
-        super().__init__()
+        super().__init__(config)
 
         self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.use_bias)
         self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.use_bias)

src/transformers/models/esm/openfold_utils/residue_constants.py

@@ -541,7 +541,6 @@ def make_bond_key(atom1_name: str, atom2_name: str) -> str:
 
 # A compact atom encoding with 14 columns
 # pylint: disable=line-too-long
-# pylint: disable=bad-whitespace
 restype_name_to_atom14_names: dict[str, list[str]] = {
     "ALA": ["N", "CA", "C", "O", "CB", "", "", "", "", "", "", "", "", ""],
     "ARG": ["N", "CA", "C", "O", "CB", "CG", "CD", "NE", "CZ", "NH1", "NH2", "", "", ""],
@@ -566,7 +565,6 @@ def make_bond_key(atom1_name: str, atom2_name: str) -> str:
     "UNK": ["", "", "", "", "", "", "", "", "", "", "", "", "", ""],
 }
 # pylint: enable=line-too-long
-# pylint: enable=bad-whitespace
 
 
 # This is the standard residue order when coding AA type as a number.

src/transformers/models/evolla/modular_evolla.py

@@ -65,7 +65,7 @@
 
 class EvollaSaProtEmbeddings(EsmEmbeddings):
     def __init__(self, config):
-        super().__init__()
+        super().__init__(config)
         # remove the position_ids in EsmEmbeddings
         self.position_ids = None
 
@@ -127,7 +127,7 @@ def forward(self, q: torch.Tensor, k: torch.Tensor) -> tuple[torch.Tensor, torch
         )
 
 
-class EvollaSaProtSelfAttention(EsmSelfAttention, nn.Module):
+class EvollaSaProtSelfAttention(EsmSelfAttention):
     def __init__(self, config, position_embedding_type=None, layer_idx=None, is_cross_attention=False):
         nn.Module.__init__(self)
         self.config = config

src/transformers/models/falcon_h1/modeling_falcon_h1.py

@@ -1007,7 +1007,7 @@ def forward(
 
 
 class FalconH1MLP(nn.Module):
-    def __init__(self, config: FalconH1Config = None):
+    def __init__(self, config: FalconH1Config):
         super().__init__()
         self.config = config
         self.hidden_size = config.hidden_size