#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
#           This file was automatically generated from src/transformers/models/lfm3/modular_lfm3.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_lfm3.py file directly. One of our CI enforces this.
#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# Copyright 2026 The HuggingFace 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 huggingface_hub.dataclasses import strict

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


@auto_docstring(checkpoint="LiquidAI/LFM3-1.7B")
@strict
class Lfm3Config(PreTrainedConfig):
    r"""
    conv_bias (`bool`, *optional*, defaults to `False`):
        Whether to use bias in the conv layers.
    conv_L_cache (`int`, *optional*, defaults to 7):
        L_cache dim in the conv layers.
    use_sandwich_norm (`bool`, *optional*, defaults to `True`):
        Whether to apply an RMSNorm to the operator and feed-forward outputs before they enter the residual stream.
    use_attention_gate (`bool`, *optional*, defaults to `True`):
        Whether the attention layers scale each head's output by a sigmoid gate computed from the layer input.

    ```python
    >>> from transformers import Lfm3Model, Lfm3Config

    >>> # Initializing a LFM3 model
    >>> configuration = Lfm3Config()

    >>> # Initializing a model from the LFM3-1.7B style configuration
    >>> model = Lfm3Model(configuration)

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

    model_type = "lfm3"
    keys_to_ignore_at_inference = ["past_key_values"]
    default_theta = 1000000.0

    vocab_size: int = 128_000
    hidden_size: int = 2048
    intermediate_size: int = 7680
    num_hidden_layers: int = 24
    num_attention_heads: int = 16
    num_key_value_heads: int = 4
    max_position_embeddings: int = 128_000
    initializer_range: float = 0.02
    norm_eps: float = 0.00001
    use_cache: bool = True
    pad_token_id: int | None = 124_893
    bos_token_id: int | None = 124_894
    eos_token_id: int | list[int] | None = 124_900
    tie_word_embeddings: bool = True
    rope_parameters: dict | None = None
    conv_bias: bool = False
    conv_L_cache: int = 7
    layer_types: list[str] | None = None
    attribute_map = {"rms_norm_eps": "norm_eps"}
    _rope_type_labels = ("full_attention", "sliding_attention")
    head_dim: int = 128
    sliding_window: int | None = 256
    use_sandwich_norm: bool = True
    use_attention_gate: bool = True
    attention_bias: bool = False
    attention_dropout: float = 0.0

    def __post_init__(self, **kwargs):
        if self.rope_parameters is None:
            self.rope_parameters = {
                "sliding_attention": {"rope_type": "default", "rope_theta": 10_000.0},
                "full_attention": {
                    "rope_type": "proportional",
                    "rope_theta": 1_000_000.0,
                    "partial_rotary_factor": 0.5,
                },
            }
        if self.layer_types is None:
            self.layer_types = [
                "full_attention" if i % 4 == 2 else "sliding_attention" if i % 4 == 0 and i > 0 else "conv"
                for i in range(self.num_hidden_layers)
            ]

        super().__post_init__(**kwargs)


__all__ = ["Lfm3Config"]
