🚨 Validate config attributes

setup.py

@@ -87,7 +87,7 @@
     "fugashi>=1.0",
     "GitPython<3.1.19",
     "hf-doc-builder>=0.3.0",
-    "huggingface-hub>=1.3.0,<2.0",
+    "huggingface-hub>=1.4.0,<2.0",
     "importlib_metadata",
     "ipadic>=1.0.0,<2.0",
     "jinja2>=3.1.0",

src/transformers/dependency_versions_table.py

@@ -19,7 +19,7 @@
     "fugashi": "fugashi>=1.0",
     "GitPython": "GitPython<3.1.19",
     "hf-doc-builder": "hf-doc-builder>=0.3.0",
-    "huggingface-hub": "huggingface-hub>=1.3.0,<2.0",
+    "huggingface-hub": "huggingface-hub>=1.4.0,<2.0",
     "importlib_metadata": "importlib_metadata",
     "ipadic": "ipadic>=1.0.0,<2.0",
     "jinja2": "jinja2>=3.1.0",

src/transformers/modeling_rope_utils.py

@@ -642,7 +642,7 @@ def convert_rope_params_to_dict(self, ignore_keys_at_rope_validation: set | None
             ignore_keys_at_rope_validation = {"partial_rotary_factor"}
 
         self.standardize_rope_params()
-        self.validate_rope(ignore_keys=ignore_keys_at_rope_validation)
+        self._validate_rope(ignore_keys=ignore_keys_at_rope_validation)
         return kwargs
 
     def standardize_rope_params(self):
@@ -693,7 +693,7 @@ def standardize_rope_params(self):
 
         self.rope_parameters = rope_parameters
 
-    def validate_rope(self: "PreTrainedConfig", ignore_keys: set | None = None):
+    def _validate_rope(self: "PreTrainedConfig", ignore_keys: set | None = None):
         """
         Validate the RoPE config arguments, given a `"PreTrainedConfig"` object
         """
@@ -926,11 +926,11 @@ def rope_config_validation(config: RotaryEmbeddingConfigMixin, ignore_keys: set
     """
     warnings.warn(
         "`rope_config_validation` is deprecated and has been removed. "
-        "Its functionality has been moved to RotaryEmbeddingConfigMixin.validate_rope method. "
-        "PreTrainedConfig inherits this class, so please call self.validate_rope() instead. "
+        "Its functionality has been moved to RotaryEmbeddingConfigMixin._validate_rope method. "
+        "PreTrainedConfig inherits this class, so please call self._validate_rope() instead. "
         "Also, make sure to use the new rope_parameters syntax. "
         "You can call self.standardize_rope_params() in the meantime.",
         FutureWarning,
     )
     config.standardize_rope_params()
-    config.validate_rope(ignore_keys=ignore_keys)
+    config._validate_rope(ignore_keys=ignore_keys)

src/transformers/models/bart/configuration_bart.py

@@ -13,13 +13,19 @@
 # limitations under the License.
 """BART model configuration"""
 
+from dataclasses import dataclass
+
+from huggingface_hub.dataclasses import strict
+
 from ...configuration_utils import PreTrainedConfig
 from ...utils import logging
 
 
 logger = logging.get_logger(__name__)
 
 
+@strict(accept_kwargs=True)
+@dataclass(repr=False)
 class BartConfig(PreTrainedConfig):
     r"""
     This is the configuration class to store the configuration of a [`BartModel`]. It is used to instantiate a BART
@@ -97,70 +103,38 @@ class BartConfig(PreTrainedConfig):
     keys_to_ignore_at_inference = ["past_key_values"]
     attribute_map = {"num_attention_heads": "encoder_attention_heads", "hidden_size": "d_model"}
 
-    def __init__(
-        self,
-        vocab_size=50265,
-        max_position_embeddings=1024,
-        encoder_layers=12,
-        encoder_ffn_dim=4096,
-        encoder_attention_heads=16,
-        decoder_layers=12,
-        decoder_ffn_dim=4096,
-        decoder_attention_heads=16,
-        encoder_layerdrop=0.0,
-        decoder_layerdrop=0.0,
-        activation_function="gelu",
-        d_model=1024,
-        dropout=0.1,
-        attention_dropout=0.0,
-        activation_dropout=0.0,
-        init_std=0.02,
-        classifier_dropout=0.0,
-        scale_embedding=False,
-        use_cache=True,
-        num_labels=3,
-        pad_token_id=1,
-        bos_token_id=0,
-        eos_token_id=2,
-        is_encoder_decoder=True,
-        decoder_start_token_id=2,
-        forced_eos_token_id=2,
-        is_decoder=False,
-        tie_word_embeddings=True,
-        **kwargs,
-    ):
-        self.is_decoder = is_decoder
-        self.tie_word_embeddings = tie_word_embeddings
-        self.vocab_size = vocab_size
-        self.max_position_embeddings = max_position_embeddings
-        self.d_model = d_model
-        self.encoder_ffn_dim = encoder_ffn_dim
-        self.encoder_layers = encoder_layers
-        self.encoder_attention_heads = encoder_attention_heads
-        self.decoder_ffn_dim = decoder_ffn_dim
-        self.decoder_layers = decoder_layers
-        self.decoder_attention_heads = decoder_attention_heads
-        self.dropout = dropout
-        self.attention_dropout = attention_dropout
-        self.activation_dropout = activation_dropout
-        self.activation_function = activation_function
-        self.init_std = init_std
-        self.encoder_layerdrop = encoder_layerdrop
-        self.decoder_layerdrop = decoder_layerdrop
-        self.classifier_dropout = classifier_dropout
-        self.use_cache = use_cache
-        self.num_hidden_layers = encoder_layers
-        self.scale_embedding = scale_embedding  # scale factor will be sqrt(d_model) if True
-
-        self.pad_token_id = pad_token_id
-        self.bos_token_id = bos_token_id
-        self.eos_token_id = eos_token_id
-        self.decoder_start_token_id = decoder_start_token_id
-        super().__init__(
-            num_labels=num_labels,
-            is_encoder_decoder=is_encoder_decoder,
-            **kwargs,
-        )
+    vocab_size: int | None = 50265
+    max_position_embeddings: int | None = 1024
+    encoder_layers: int | None = 12
+    encoder_ffn_dim: int | None = 4096
+    encoder_attention_heads: int | None = 16
+    decoder_layers: int | None = 12
+    decoder_ffn_dim: int | None = 4096
+    decoder_attention_heads: int | None = 16
+    encoder_layerdrop: float | None = 0.0
+    decoder_layerdrop: float | None = 0.0
+    activation_function: str | None = "gelu"
+    d_model: int | None = 1024
+    dropout: float | None = 0.1
+    attention_dropout: float | None = 0.0
+    activation_dropout: float | None = 0.0
+    init_std: float | None = 0.02
+    classifier_dropout: float | None = 0.0
+    scale_embedding: bool | None = False
+    use_cache: bool | None = True
+    pad_token_id: int | None = 1
+    bos_token_id: int | None = 0
+    eos_token_id: int | None = 2
+    is_encoder_decoder: bool | None = True
+    decoder_start_token_id: int | None = 2
+    forced_eos_token_id: int | None = 2
+    is_decoder: bool | None = False
+    tie_word_embeddings: bool | None = True
+    num_labels: int | None = 3
+
+    def __post_init__(self, **kwargs):
+        self.num_hidden_layers = self.encoder_layers
+        super().__post_init__(**kwargs)
 
 
 __all__ = ["BartConfig"]