Make XLMRoberta model and config independent from Roberta (#19359)

* remove config dependence

* remove dependencies from xlm_roberta

* Fix style

* Fix comments

* various fixes

* Fix pre-trained model name

src/transformers/__init__.py

@@ -2125,6 +2125,7 @@ else:
             "XLMRobertaForSequenceClassification",
             "XLMRobertaForTokenClassification",
             "XLMRobertaModel",
+            "XLMRobertaPreTrainedModel",
         ]
     )
     _import_structure["models.xlm_roberta_xl"].extend(
@@ -4805,6 +4806,7 @@ if TYPE_CHECKING:
             XLMRobertaForSequenceClassification,
             XLMRobertaForTokenClassification,
             XLMRobertaModel,
+            XLMRobertaPreTrainedModel,
         )
         from .models.xlm_roberta_xl import (
             XLM_ROBERTA_XL_PRETRAINED_MODEL_ARCHIVE_LIST,

src/transformers/models/xlm_roberta/__init__.py

@@ -68,6 +68,7 @@ else:
         "XLMRobertaForSequenceClassification",
         "XLMRobertaForTokenClassification",
         "XLMRobertaModel",
+        "XLMRobertaPreTrainedModel",
     ]
 
 try:
@@ -139,6 +140,7 @@ if TYPE_CHECKING:
             XLMRobertaForSequenceClassification,
             XLMRobertaForTokenClassification,
             XLMRobertaModel,
+            XLMRobertaPreTrainedModel,
         )
 
     try:

src/transformers/models/xlm_roberta/configuration_xlm_roberta.py

@@ -17,9 +17,9 @@
 from collections import OrderedDict
 from typing import Mapping
 
+from ...configuration_utils import PretrainedConfig
 from ...onnx import OnnxConfig
 from ...utils import logging
-from ..roberta.configuration_roberta import RobertaConfig
 
 
 logger = logging.get_logger(__name__)
@@ -42,15 +42,114 @@ XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP = {
 }
 
 
-class XLMRobertaConfig(RobertaConfig):
-    """
-    This class overrides [`RobertaConfig`]. Please check the superclass for the appropriate documentation alongside
-    usage examples. Instantiating a configuration with the defaults will yield a similar configuration to that of the
-    XLMRoBERTa [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) architecture.
-    """
+class XLMRobertaConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`XLMRobertaModel`] or a [`TFXLMRobertaModel`]. It
+    is used to instantiate a XLM-RoBERTa model according to the specified arguments, defining the model architecture.
+    Instantiating a configuration with the defaults will yield a similar configuration to that of the XLMRoBERTa
+    [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) architecture.
 
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 30522):
+            Vocabulary size of the XLM-RoBERTa model. Defines the number of different tokens that can be represented by
+            the `inputs_ids` passed when calling [`XLMRobertaModel`] or [`TFXLMRobertaModel`].
+        hidden_size (`int`, *optional*, defaults to 768):
+            Dimensionality of the encoder layers and the pooler layer.
+        num_hidden_layers (`int`, *optional*, defaults to 12):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 12):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        intermediate_size (`int`, *optional*, defaults to 3072):
+            Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
+        hidden_act (`str` or `Callable`, *optional*, defaults to `"gelu"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"silu"` and `"gelu_new"` are supported.
+        hidden_dropout_prob (`float`, *optional*, defaults to 0.1):
+            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
+        attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):
+            The dropout ratio for the attention probabilities.
+        max_position_embeddings (`int`, *optional*, defaults to 512):
+            The maximum sequence length that this model might ever be used with. Typically set this to something large
+            just in case (e.g., 512 or 1024 or 2048).
+        type_vocab_size (`int`, *optional*, defaults to 2):
+            The vocabulary size of the `token_type_ids` passed when calling [`XLMRobertaModel`] or
+            [`TFXLMRobertaModel`].
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-12):
+            The epsilon used by the layer normalization layers.
+        position_embedding_type (`str`, *optional*, defaults to `"absolute"`):
+            Type of position embedding. Choose one of `"absolute"`, `"relative_key"`, `"relative_key_query"`. For
+            positional embeddings use `"absolute"`. For more information on `"relative_key"`, please refer to
+            [Self-Attention with Relative Position Representations (Shaw et al.)](https://arxiv.org/abs/1803.02155).
+            For more information on `"relative_key_query"`, please refer to *Method 4* in [Improve Transformer Models
+            with Better Relative Position Embeddings (Huang et al.)](https://arxiv.org/abs/2009.13658).
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        classifier_dropout (`float`, *optional*):
+            The dropout ratio for the classification head.
+
+    Examples:
+
+    ```python
+    >>> from transformers import XLMRobertaModel, XLMRobertaConfig
+
+    >>> # Initializing a XLM-RoBERTa xlm-roberta-base style configuration
+    >>> configuration = XLMRobertaConfig()
+
+    >>> # Initializing a model from the xlm-roberta-base style configuration
+    >>> model = XLMRobertaModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
     model_type = "xlm-roberta"
 
+    def __init__(
+        self,
+        vocab_size=30522,
+        hidden_size=768,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+        intermediate_size=3072,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=512,
+        type_vocab_size=2,
+        initializer_range=0.02,
+        layer_norm_eps=1e-12,
+        pad_token_id=1,
+        bos_token_id=0,
+        eos_token_id=2,
+        position_embedding_type="absolute",
+        use_cache=True,
+        classifier_dropout=None,
+        **kwargs
+    ):
+        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
+
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.hidden_act = hidden_act
+        self.intermediate_size = intermediate_size
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.type_vocab_size = type_vocab_size
+        self.initializer_range = initializer_range
+        self.layer_norm_eps = layer_norm_eps
+        self.position_embedding_type = position_embedding_type
+        self.use_cache = use_cache
+        self.classifier_dropout = classifier_dropout
+
 
 # Copied from transformers.models.roberta.configuration_roberta.RobertaOnnxConfig with Roberta->XLMRoberta
 class XLMRobertaOnnxConfig(OnnxConfig):

src/transformers/utils/dummy_pt_objects.py

@@ -5689,6 +5689,13 @@ class XLMRobertaModel(metaclass=DummyObject):
         requires_backends(self, ["torch"])
 
 
+class XLMRobertaPreTrainedModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
 XLM_ROBERTA_XL_PRETRAINED_MODEL_ARCHIVE_LIST = None