#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
#           This file was automatically generated from src/transformers/models/modernbert/modular_modernbert.py.
#               Do NOT edit this file manually as any edits will be overwritten by the generation of
#             the file from the modular. If any change should be done, please apply the change to the
#                          modular_modernbert.py file directly. One of our CI enforces this.
#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# Copyright 2024 Answer.AI, LightOn, and contributors, and the HuggingFace Inc. team. All rights reserved.
#
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from typing import Literal

from huggingface_hub.dataclasses import strict

from ...configuration_utils import PreTrainedConfig
from ...utils import auto_docstring


@auto_docstring(checkpoint="answerdotai/ModernBERT-base")
@strict
class ModernBertConfig(PreTrainedConfig):
    r"""
    initializer_cutoff_factor (`float`, *optional*, defaults to 2.0):
        The cutoff factor for the truncated_normal_initializer for initializing all weight matrices.
    norm_eps (`float`, *optional*, defaults to 1e-05):
        The epsilon used by the rms normalization layers.
    norm_bias (`bool`, *optional*, defaults to `False`):
        Whether to use bias in the normalization layers.
    local_attention (`int`, *optional*, defaults to 128):
        The window size for local attention.
    mlp_dropout (`float`, *optional*, defaults to 0.0):
        The dropout ratio for the MLP layers.
    decoder_bias (`bool`, *optional*, defaults to `True`):
        Whether to use bias in the decoder layers.
    classifier_pooling (`str`, *optional*, defaults to `"cls"`):
        The pooling method for the classifier. Should be either `"cls"` or `"mean"`. In local attention layers, the
        CLS token doesn't attend to all tokens on long sequences.
    classifier_bias (`bool`, *optional*, defaults to `False`):
        Whether to use bias in the classifier.
    classifier_activation (`str`, *optional*, defaults to `"gelu"`):
        The activation function for the classifier.
    deterministic_flash_attn (`bool`, *optional*, defaults to `False`):
        Whether to use deterministic flash attention. If `False`, inference will be faster but not deterministic.
    sparse_prediction (`bool`, *optional*, defaults to `False`):
        Whether to use sparse prediction for the masked language model instead of returning the full dense logits.
    sparse_pred_ignore_index (`int`, *optional*, defaults to -100):
        The index to ignore for the sparse prediction.

    Examples:

    ```python
    >>> from transformers import ModernBertModel, ModernBertConfig

    >>> # Initializing a ModernBert style configuration
    >>> configuration = ModernBertConfig()

    >>> # Initializing a model from the modernbert-base style configuration
    >>> model = ModernBertModel(configuration)

    >>> # Accessing the model configuration
    >>> configuration = model.config
    ```"""

    model_type = "modernbert"
    keys_to_ignore_at_inference = ["past_key_values"]
    default_theta = {"global": 160_000.0, "local": 10_000.0}

    vocab_size: int = 50368
    hidden_size: int = 768
    intermediate_size: int = 1152
    num_hidden_layers: int = 22
    num_attention_heads: int = 12
    hidden_activation: str = "gelu"
    max_position_embeddings: int = 8192
    initializer_range: float = 0.02
    initializer_cutoff_factor: float = 2.0
    norm_eps: float = 1e-5
    norm_bias: bool = False
    pad_token_id: int | None = 50283
    eos_token_id: int | list[int] | None = 50282
    bos_token_id: int | None = 50281
    cls_token_id: int | None = 50281
    sep_token_id: int | None = 50282
    attention_bias: bool = False
    attention_dropout: float | int = 0.0
    layer_types: list[str] | None = None
    rope_parameters: dict[Literal["full_attention", "sliding_attention"], dict] | None = None
    local_attention: int = 128
    embedding_dropout: float | int = 0.0
    mlp_bias: bool = False
    mlp_dropout: float | int = 0.0
    decoder_bias: bool = True
    classifier_pooling: Literal["cls", "mean"] = "cls"
    classifier_dropout: float | int = 0.0
    classifier_bias: bool = False
    classifier_activation: str = "gelu"
    deterministic_flash_attn: bool = False
    sparse_prediction: bool = False
    sparse_pred_ignore_index: int = -100
    tie_word_embeddings: bool = True

    def __post_init__(self, **kwargs):
        # BC -> the pattern used to be a simple int, and it's still present in configs on the Hub
        global_attn_every_n_layers = kwargs.get("global_attn_every_n_layers", 3)
        if self.layer_types is None:
            self.layer_types = [
                "sliding_attention" if bool(i % global_attn_every_n_layers) else "full_attention"
                for i in range(self.num_hidden_layers)
            ]

        super().__post_init__(**kwargs)

    def convert_rope_params_to_dict(self, **kwargs):
        rope_scaling = kwargs.pop("rope_scaling", None)

        # Try to set `rope_scaling` if available, otherwise use `rope_parameters`. If we find `rope_parameters`
        # as arg in the inputs, we can safely assume that it is in the new format. New naming used -> new format
        default_rope_params = {
            "sliding_attention": {"rope_type": "default"},
            "full_attention": {"rope_type": "default"},
        }
        self.rope_parameters = self.rope_parameters if self.rope_parameters is not None else default_rope_params
        if rope_scaling is not None:
            self.rope_parameters["full_attention"].update(rope_scaling)
            self.rope_parameters["sliding_attention"].update(rope_scaling)

        # Set default values if not present
        if self.rope_parameters.get("full_attention") is None:
            self.rope_parameters["full_attention"] = {"rope_type": "default"}
        self.rope_parameters["full_attention"].setdefault(
            "rope_theta", kwargs.pop("global_rope_theta", self.default_theta["global"])
        )
        if self.rope_parameters.get("sliding_attention") is None:
            self.rope_parameters["sliding_attention"] = {"rope_type": "default"}
        self.rope_parameters["sliding_attention"].setdefault(
            "rope_theta", kwargs.pop("local_rope_theta", self.default_theta["local"])
        )

        # Standardize and validate the correctness of rotary position embeddings parameters
        self.standardize_rope_params()
        return kwargs

    def to_dict(self):
        output = super().to_dict()
        output.pop("reference_compile", None)
        return output

    @property
    def sliding_window(self):
        """Half-window size: `local_attention` is the total window, so we divide by 2."""
        return self.local_attention // 2

    @sliding_window.setter
    def sliding_window(self, value):
        """Set sliding_window by updating local_attention to 2 * value."""
        self.local_attention = value * 2


__all__ = ["ModernBertConfig"]