Config annotations (#50067)

* clean against TRF 055

* and the rest

* auto fix

* i fixed it no? Where it went fix again

* some more, still needs to fix repo

* sync reverse

* and these

* make the TRF stop compaining

* esm was nevr using the weight init...

* forgot

* what is the point of unused class in that case?

Author: zucchini-nlp

src/transformers/models/afmoe/modeling_afmoe.py

@@ -594,7 +594,7 @@ class AfmoeForCausalLM(AfmoePreTrainedModel, GenerationMixin):
 
     _fsdp_plan = {"lm_head": "keep_full_weight"}
 
-    def __init__(self, config):
+    def __init__(self, config: AfmoeConfig):
         super().__init__(config)
         self.model = AfmoeModel(config)
         self.vocab_size = config.vocab_size

src/transformers/models/afmoe/modular_afmoe.py

@@ -398,7 +398,7 @@ class AfmoeForCausalLM(LlamaForCausalLM, AfmoePreTrainedModel, GenerationMixin):
 
     _fsdp_plan = {"lm_head": "keep_full_weight"}
 
-    def __init__(self, config):
+    def __init__(self, config: AfmoeConfig):
         AfmoePreTrainedModel.__init__(self, config)
         self.model = AfmoeModel(config)
         self.vocab_size = config.vocab_size

src/transformers/models/aimv2/modeling_aimv2.py

@@ -535,7 +535,7 @@ def forward(
 )
 class Aimv2TextModel(Aimv2PreTrainedModel):
     main_input_name = "input_ids"
-
+    config: Aimv2TextConfig
     _can_record_outputs = {
         "hidden_states": Aimv2EncoderLayer,
         "attentions": Aimv2Attention,

src/transformers/models/aimv2/modular_aimv2.py

@@ -386,7 +386,7 @@ def forward(
 )
 class Aimv2TextModel(Aimv2PreTrainedModel):
     main_input_name = "input_ids"
-
+    config: Aimv2TextConfig
     _can_record_outputs = {
         "hidden_states": Aimv2EncoderLayer,
         "attentions": Aimv2Attention,

src/transformers/models/albert/modeling_albert.py

@@ -550,7 +550,7 @@ class AlbertForMaskedLM(AlbertPreTrainedModel):
         "predictions.decoder.bias": "predictions.bias",
     }
 
-    def __init__(self, config):
+    def __init__(self, config: AlbertConfig):
         super().__init__(config)
 
         self.albert = AlbertModel(config, add_pooling_layer=False)

src/transformers/models/altclip/modeling_altclip.py

@@ -763,7 +763,7 @@ class AltRobertaModel(AltCLIPPreTrainedModel):
         "attentions": AltRobertaSelfAttention,
     }
 
-    def __init__(self, config, add_pooling_layer=True):
+    def __init__(self, config: AltCLIPTextConfig, add_pooling_layer=True):
         r"""
         add_pooling_layer (bool, *optional*, defaults to `True`):
             Whether to add a pooling layer
@@ -838,7 +838,7 @@ class AltCLIPTextModel(AltCLIPPreTrainedModel):
     _input_embed_layer = "word_embeddings"
     base_model_prefix = "roberta"
 
-    def __init__(self, config):
+    def __init__(self, config: AltCLIPTextConfig):
         super().__init__(config)
         self.roberta = AltRobertaModel(config, add_pooling_layer=False)
         self.transformation = nn.Linear(config.hidden_size, config.project_dim)

src/transformers/models/altclip/modular_altclip.py

@@ -315,7 +315,7 @@ class AltRobertaModel(AltCLIPPreTrainedModel):
         "attentions": AltRobertaSelfAttention,
     }
 
-    def __init__(self, config, add_pooling_layer=True):
+    def __init__(self, config: AltCLIPTextConfig, add_pooling_layer=True):
         r"""
         add_pooling_layer (bool, *optional*, defaults to `True`):
             Whether to add a pooling layer
@@ -390,7 +390,7 @@ class AltCLIPTextModel(AltCLIPPreTrainedModel):
     _input_embed_layer = "word_embeddings"
     base_model_prefix = "roberta"
 
-    def __init__(self, config):
+    def __init__(self, config: AltCLIPTextConfig):
         super().__init__(config)
         self.roberta = AltRobertaModel(config, add_pooling_layer=False)
         self.transformation = nn.Linear(config.hidden_size, config.project_dim)

src/transformers/models/apertus/modeling_apertus.py

@@ -407,7 +407,7 @@ class ApertusForCausalLM(ApertusPreTrainedModel, GenerationMixin):
     _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
     _fsdp_plan = {"lm_head": "keep_full_weight"}
 
-    def __init__(self, config):
+    def __init__(self, config: ApertusConfig):
         super().__init__(config)
         self.model = ApertusModel(config)
         self.vocab_size = config.vocab_size

src/transformers/models/arcee/modeling_arcee.py

@@ -409,7 +409,7 @@ class ArceeForCausalLM(ArceePreTrainedModel, GenerationMixin):
     _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
     _fsdp_plan = {"lm_head": "keep_full_weight"}
 
-    def __init__(self, config):
+    def __init__(self, config: ArceeConfig):
         super().__init__(config)
         self.model = ArceeModel(config)
         self.vocab_size = config.vocab_size

src/transformers/models/audioflamingo3/modeling_audioflamingo3.py

@@ -446,7 +446,7 @@ class AudioFlamingo3Model(AudioFlamingo3PreTrainedModel):
     _pp_plan = None
     _keep_in_fp32_modules_strict = None
 
-    def __init__(self, config):
+    def __init__(self, config: AudioFlamingo3Config):
         super().__init__(config)
         self.audio_tower = AutoModel.from_config(config.audio_config)
         self.language_model = AutoModel.from_config(config.text_config)
@@ -566,7 +566,7 @@ def forward(
 class AudioFlamingo3ForConditionalGeneration(AudioFlamingo3PreTrainedModel, GenerationMixin):
     _tied_weights_keys = None
 
-    def __init__(self, config):
+    def __init__(self, config: AudioFlamingo3Config):
         super().__init__(config)
         self.model = AudioFlamingo3Model(config)
         self.lm_head = nn.Linear(config.text_config.hidden_size, config.text_config.vocab_size, bias=False)