#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
#           This file was automatically generated from src/transformers/models/esmc/modular_esmc.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_esmc.py file directly. One of our CI enforces this.
#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# Copyright 2026 BioHub 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 huggingface_hub.dataclasses import strict

from ...configuration_utils import PreTrainedConfig
from ...modeling_rope_utils import RopeParameters
from ...utils import auto_docstring
from ...utils.type_validators import interval


@auto_docstring(checkpoint="biohub/ESMC-6B-hf")
@strict
class EsmcConfig(PreTrainedConfig):
    r"""
    mask_token_id (`int`, *optional*, defaults to 32):
        Index of the mask token in the vocabulary (``"<mask>"``), used for masked language modelling.
    classifier_dropout (`float`, *optional*, defaults to 0.1):
        Dropout ratio for the classification head.

    Examples:

    ```python
    >>> from transformers import EsmcConfig, EsmcModel

    >>> # Initializing an ESMC biohub/ESMC-6B-hf style configuration
    >>> configuration = EsmcConfig()

    >>> # Initializing a model (with random weights) from the configuration
    >>> model = EsmcModel(configuration)

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

    model_type = "esmc"
    # Default tensor parallel plan for base model `EsmcModel`
    base_model_tp_plan = {
        "layers.*.self_attn.q_proj": "colwise",
        "layers.*.self_attn.k_proj": "colwise",
        "layers.*.self_attn.v_proj": "colwise",
        "layers.*.self_attn.o_proj": "rowwise",
        "layers.*.mlp.gate_proj": "colwise",
        "layers.*.mlp.up_proj": "colwise",
        "layers.*.mlp.down_proj": "rowwise",
    }
    base_model_pp_plan = {
        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
        "norm": (["hidden_states"], ["hidden_states"]),
    }

    # Llama fields re-declared only where ESMC's default differs from the parent's.
    vocab_size: int = 64
    hidden_size: int = 2560
    intermediate_size: int = 6912
    num_hidden_layers: int = 80
    num_attention_heads: int = 40
    num_key_value_heads: int | None = None
    hidden_act: str = "silu"
    max_position_embeddings: int = 2048
    initializer_range: float = interval(min=0.0, max=1.0)(default=0.02)
    pad_token_id: int | None = 1
    bos_token_id: int | None = 0
    eos_token_id: int | list[int] | None = 2
    tie_word_embeddings: bool = False
    rope_parameters: RopeParameters | dict | None = None
    attention_bias: bool = False
    attention_dropout: int | float | None = 0.0
    mlp_bias: bool = False
    head_dim: int | None = None

    # ESMC-specific fields.
    mask_token_id: int | None = 32
    classifier_dropout: float | None = 0.1

    def __post_init__(self, **kwargs):
        # The special-token ids are fixed by the vocabulary every checkpoint shares (`<cls>`=0 doubles
        # as BOS, `<eos>`=2); configs saved before these fields existed carry explicit nulls.
        if self.bos_token_id is None:
            self.bos_token_id = 0
        if self.eos_token_id is None:
            self.eos_token_id = 2
        if self.head_dim is None:
            self.head_dim = self.hidden_size // self.num_attention_heads
        if self.num_key_value_heads is None:
            self.num_key_value_heads = self.num_attention_heads

        super().__post_init__(**kwargs)

    def validate_architecture(self):
        """Part of `@strict`-powered validation. Validates the architecture of the config."""
        if self.hidden_size % self.num_attention_heads != 0:
            raise ValueError(
                f"The hidden size ({self.hidden_size}) is not a multiple of the number of attention "
                f"heads ({self.num_attention_heads})."
            )
        # ESMC never uses grouped-query attention; the parent derives the two to match when unset.
        if self.num_key_value_heads != self.num_attention_heads:
            raise ValueError(
                f"ESMC does not support grouped-query attention: `num_key_value_heads` "
                f"({self.num_key_value_heads}) must equal `num_attention_heads` ({self.num_attention_heads})."
            )


__all__ = ["EsmcConfig"]