#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
#           This file was automatically generated from src/transformers/models/lighton_ocr/modular_lighton_ocr.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_lighton_ocr.py file directly. One of our CI enforces this.
#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# Copyright 2026 The LightOn Team 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 dataclasses import dataclass

import torch
from torch import nn

from ...cache_utils import Cache
from ...generation import GenerationMixin
from ...integrations import use_kernel_forward_from_hub
from ...modeling_outputs import BaseModelOutputWithPast, BaseModelOutputWithPooling, ModelOutput
from ...modeling_utils import PreTrainedModel
from ...processing_utils import Unpack
from ...utils import TransformersKwargs, auto_docstring, can_return_tuple, torch_compilable_check
from ..auto import AutoModel
from .configuration_lighton_ocr import LightOnOcrConfig


@use_kernel_forward_from_hub("RMSNorm")
class LightOnOcrRMSNorm(nn.Module):
    def __init__(self, hidden_size, eps: float = 1e-6) -> None:
        """
        LightOnOcrRMSNorm is equivalent to T5LayerNorm
        """
        super().__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size))
        self.variance_epsilon = eps

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        input_dtype = hidden_states.dtype
        hidden_states = hidden_states.to(torch.float32)
        variance = hidden_states.pow(2).mean(-1, keepdim=True)
        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
        return self.weight * hidden_states.to(input_dtype)

    def extra_repr(self):
        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"


class LightOnOcrPatchMerger(nn.Module):
    """
    Learned merging of spatial_merge_size ** 2 patches
    """

    def __init__(self, config: LightOnOcrConfig):
        super().__init__()
        self.config = config

        hidden_size = config.vision_config.hidden_size
        self.spatial_merge_size = config.spatial_merge_size
        self.patch_size = self.config.vision_config.patch_size
        self.merging_layer = nn.Linear(hidden_size * self.spatial_merge_size**2, hidden_size, bias=False)

    def forward(self, image_features: torch.Tensor, image_sizes: torch.Tensor) -> torch.Tensor:
        image_sizes = [
            (image_size[0] // self.patch_size, image_size[1] // self.patch_size) for image_size in image_sizes
        ]

        tokens_per_image = [h * w for h, w in image_sizes]
        d = image_features.shape[-1]

        permuted_tensor = []
        for image_index, image_tokens in enumerate(image_features.split(tokens_per_image)):
            # Reshape image_tokens into a 2D grid
            h, w = image_sizes[image_index]
            image_grid = image_tokens.view(h, w, d).permute(2, 0, 1).unsqueeze(0)
            grid = torch.nn.functional.unfold(
                image_grid, kernel_size=self.spatial_merge_size, stride=self.spatial_merge_size
            )
            grid = grid.view(d * self.spatial_merge_size**2, -1).t()
            permuted_tensor.append(grid)

        image_features = torch.cat(permuted_tensor, dim=0)
        image_features = self.merging_layer(image_features)
        return image_features


class LightOnOcrMultiModalProjector(nn.Module):
    def __init__(self, config: LightOnOcrConfig):
        super().__init__()
        self.config = config
        self.norm = LightOnOcrRMSNorm(config.vision_config.hidden_size, eps=config.text_config.rms_norm_eps)
        self.patch_merger = LightOnOcrPatchMerger(config)
        self.linear_1 = nn.Linear(config.vision_config.hidden_size, config.text_config.hidden_size, bias=False)
        self.act = nn.GELU()
        self.linear_2 = nn.Linear(config.text_config.hidden_size, config.text_config.hidden_size, bias=False)

    def forward(self, image_features: torch.Tensor, image_sizes: torch.Tensor):
        image_features = self.norm(image_features)
        image_features = self.patch_merger(image_features, image_sizes)
        hidden_states = self.linear_1(image_features)
        hidden_states = self.act(hidden_states)
        hidden_states = self.linear_2(hidden_states)
        return hidden_states


@dataclass
@auto_docstring(
    custom_intro="""
    Base class for LightOnOcr outputs, with hidden states and attentions.
    """
)
class LightOnOcrModelOutputWithPast(BaseModelOutputWithPast):
    r"""
    past_key_values (`Cache`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
        It is a [`~cache_utils.Cache`] instance. For more details, see our [kv cache guide](https://huggingface.co/docs/transformers/en/kv_cache).

        Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see
        `past_key_values` input) to speed up sequential decoding.
    image_hidden_states (`torch.FloatTensor`, *optional*):
        A `torch.FloatTensor` of size `(batch_size, num_images, sequence_length, hidden_size)`.
        image_hidden_states of the model produced by the vision encoder and after projecting the last hidden state.
    """

    image_hidden_states: torch.FloatTensor | None = None


@auto_docstring
class LightOnOcrPreTrainedModel(PreTrainedModel):
    config: LightOnOcrConfig
    base_model_prefix = "model"
    input_modalities = ("image", "text")
    supports_gradient_checkpointing = True
    _skip_keys_device_placement = "past_key_values"

    _supports_flash_attn = True
    _supports_sdpa = True

    _can_compile_fullgraph = True
    _supports_flex_attn = True
    _supports_attention_backend = True


@auto_docstring(
    custom_intro="""
    The LightOnOcr model which consists of a vision backbone and a language model, without a language modeling head.
    """
)
class LightOnOcrModel(LightOnOcrPreTrainedModel):
    _checkpoint_conversion_mapping = {}
    base_model_prefix = ""

    def __init__(self, config: LightOnOcrConfig):
        super().__init__(config)
        self.vision_encoder = AutoModel.from_config(config.vision_config)
        self.vision_projection = LightOnOcrMultiModalProjector(config)
        self.language_model = AutoModel.from_config(config.text_config)
        self.post_init()

    def get_input_embeddings(self):
        return self.language_model.get_input_embeddings()

    def set_input_embeddings(self, value):
        self.language_model.set_input_embeddings(value)

    @can_return_tuple
    @auto_docstring
    def get_image_features(
        self, pixel_values: torch.Tensor, image_sizes: torch.Tensor | list, **kwargs: Unpack[TransformersKwargs]
    ) -> tuple | BaseModelOutputWithPooling:
        image_outputs = self.vision_encoder(pixel_values, image_sizes=image_sizes, return_dict=True, **kwargs)
        image_features = image_outputs.last_hidden_state
        image_features = self.vision_projection(image_features.squeeze(0), image_sizes)

        # Split features per image based on the effective patch size
        downsample_ratio = self.config.vision_config.patch_size * self.config.spatial_merge_size
        split_sizes = [(height // downsample_ratio) * (width // downsample_ratio) for height, width in image_sizes]
        image_features = torch.split(image_features, split_sizes)
        image_outputs.pooler_output = image_features

        return image_outputs

    def get_placeholder_mask(
        self, input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor
    ):
        """
        Obtains multimodal placeholder mask from `input_ids` or `inputs_embeds`, and checks that the placeholder token count is
        equal to the length of multimodal features. If the lengths are different, an error is raised.
        """
        if input_ids is None:
            special_image_mask = inputs_embeds == self.get_input_embeddings()(
                torch.tensor(self.config.image_token_id, dtype=torch.long, device=inputs_embeds.device)
            )
            special_image_mask = special_image_mask.all(-1)
        else:
            special_image_mask = input_ids == self.config.image_token_id

        n_image_tokens = special_image_mask.sum()
        n_image_features = image_features.shape[0] * image_features.shape[1]
        special_image_mask = special_image_mask.unsqueeze(-1).expand_as(inputs_embeds).to(inputs_embeds.device)
        torch_compilable_check(
            inputs_embeds[special_image_mask].numel() == image_features.numel(),
            f"Image features and image tokens do not match, tokens: {n_image_tokens}, features: {n_image_features}",
        )
        return special_image_mask

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.LongTensor | None = None,
        pixel_values: torch.FloatTensor | None = None,
        attention_mask: torch.Tensor | None = None,
        position_ids: torch.LongTensor | None = None,
        past_key_values: Cache | None = None,
        inputs_embeds: torch.FloatTensor | None = None,
        use_cache: bool | None = None,
        output_attentions: bool | None = None,
        output_hidden_states: bool | None = None,
        return_dict: bool | None = None,
        cache_position: torch.LongTensor | None = None,
        image_sizes: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple | LightOnOcrModelOutputWithPast:
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if (input_ids is None) ^ (inputs_embeds is not None):
            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")

        if inputs_embeds is None:
            inputs_embeds = self.get_input_embeddings()(input_ids)

        if pixel_values is not None:
            image_features = self.get_image_features(
                pixel_values=pixel_values, image_sizes=image_sizes, return_dict=True
            ).pooler_output
            image_features = torch.cat(image_features, dim=0).to(inputs_embeds.device, inputs_embeds.dtype)
            special_image_mask = self.get_placeholder_mask(
                input_ids, inputs_embeds=inputs_embeds, image_features=image_features
            )
            inputs_embeds = inputs_embeds.masked_scatter(special_image_mask, image_features)

        outputs = self.language_model(
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=True,
            cache_position=cache_position,
            **kwargs,
        )

        return LightOnOcrModelOutputWithPast(
            last_hidden_state=outputs.last_hidden_state,
            past_key_values=outputs.past_key_values,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
            image_hidden_states=image_features if pixel_values is not None else None,
        )


@dataclass
@auto_docstring(
    custom_intro="""
    Base class for LightOnOcr causal language model (or autoregressive) outputs.
    """
)
class LightOnOcrCausalLMOutputWithPast(ModelOutput):
    r"""
    loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
        Language modeling loss (for next-token prediction).
    logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`):
        Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
    past_key_values (`Cache`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
        It is a [`~cache_utils.Cache`] instance. For more details, see our [kv cache guide](https://huggingface.co/docs/transformers/en/kv_cache).

        Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see
        `past_key_values` input) to speed up sequential decoding.
    image_hidden_states (`torch.FloatTensor`, *optional*):
        A `torch.FloatTensor` of size `(batch_size, num_images, sequence_length, hidden_size)`.
        image_hidden_states of the model produced by the vision encoder and after projecting the last hidden state.
    """

    loss: torch.FloatTensor | None = None
    logits: torch.FloatTensor | None = None
    past_key_values: Cache | None = None
    hidden_states: tuple[torch.FloatTensor] | None = None
    attentions: tuple[torch.FloatTensor] | None = None
    image_hidden_states: torch.FloatTensor | None = None


@auto_docstring(
    custom_intro="""
    The LIGHTON_OCR model which consists of a vision backbone and a language model.
    """
)
class LightOnOcrForConditionalGeneration(LightOnOcrPreTrainedModel, GenerationMixin):
    _checkpoint_conversion_mapping = {}
    _tied_weights_keys = {"lm_head.weight": "model.language_model.embed_tokens.weight"}

    def __init__(self, config: LightOnOcrConfig):
        super().__init__(config)
        self.model = LightOnOcrModel(config)
        self.lm_head = nn.Linear(config.text_config.hidden_size, config.text_config.vocab_size, bias=False)
        self.post_init()

    def get_input_embeddings(self):
        return self.model.get_input_embeddings()

    def set_input_embeddings(self, value):
        self.model.set_input_embeddings(value)

    def get_output_embeddings(self) -> nn.Module:
        return self.lm_head

    @auto_docstring
    def get_image_features(
        self, pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, **kwargs: Unpack[TransformersKwargs]
    ) -> tuple | BaseModelOutputWithPooling:
        return self.model.get_image_features(pixel_values=pixel_values, image_sizes=image_sizes, **kwargs)

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.LongTensor | None = None,
        pixel_values: torch.FloatTensor | None = None,
        attention_mask: torch.Tensor | None = None,
        position_ids: torch.LongTensor | None = None,
        past_key_values: Cache | None = None,
        inputs_embeds: torch.FloatTensor | None = None,
        labels: torch.LongTensor | None = None,
        use_cache: bool | None = None,
        output_attentions: bool | None = None,
        output_hidden_states: bool | None = None,
        return_dict: bool | None = None,
        cache_position: torch.LongTensor | None = None,
        logits_to_keep: int | torch.Tensor = 0,
        image_sizes: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple | LightOnOcrCausalLMOutputWithPast:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
            config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
            (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.

        Example:

        ```python
        >>> from PIL import Image
        >>> import httpx
        >>> from io import BytesIO
        >>> from transformers import AutoProcessor, LightOnOcrForConditionalGeneration

        >>> model = LightOnOcrForConditionalGeneration.from_pretrained("mistralai/Mistral-Small-3.1-24B-Instruct-2503")
        >>> processor = AutoProcessor.from_pretrained("mistralai/Mistral-Small-3.1-24B-Instruct-2503")

        >>> prompt = "<s>[INST][IMG]What is the image?[/INST]"
        >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
        >>> with httpx.stream("GET", url) as response:
        ...     image = Image.open(BytesIO(response.read()))

        >>> inputs = processor(images=image, text=prompt, return_tensors="pt")

        >>> # Generate
        >>> generate_ids = model.generate(**inputs, max_new_tokens=15)
        >>> processor.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        "What is the image?The image depicts two cats lying on a pink blanket."
        ```"""
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.model(
            input_ids=input_ids,
            pixel_values=pixel_values,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=True,
            cache_position=cache_position,
            image_sizes=image_sizes,
            **kwargs,
        )

        hidden_states = outputs[0]
        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
        logits = self.lm_head(hidden_states[:, slice_indices, :])

        loss = None
        if labels is not None:
            loss = self.loss_function(
                logits=logits, labels=labels, vocab_size=self.config.text_config.vocab_size, **kwargs
            )

        return LightOnOcrCausalLMOutputWithPast(
            loss=loss,
            logits=logits,
            past_key_values=outputs.past_key_values,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
            image_hidden_states=outputs.image_hidden_states,
        )

    def prepare_inputs_for_generation(
        self,
        input_ids,
        past_key_values=None,
        inputs_embeds=None,
        pixel_values=None,
        attention_mask=None,
        cache_position=None,
        logits_to_keep=None,
        is_first_iteration=False,
        **kwargs,
    ):
        # Overwritten -- in specific circumstances we don't want to forward image inputs to the model

        model_inputs = super().prepare_inputs_for_generation(
            input_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            attention_mask=attention_mask,
            cache_position=cache_position,
            logits_to_keep=logits_to_keep,
            is_first_iteration=is_first_iteration,
            **kwargs,
        )

        if is_first_iteration or not kwargs.get("use_cache", True):
            # Pixel values are used only in the first iteration if available
            # In subsequent iterations, they are already merged with text and cached
            # NOTE: first iteration doesn't have to be prefill, it can be the first
            # iteration with a question and cached system prompt (continue generate from cache)
            model_inputs["pixel_values"] = pixel_values

        return model_inputs


__all__ = ["LightOnOcrPreTrainedModel", "LightOnOcrForConditionalGeneration", "LightOnOcrModel"]