# coding=utf-8
# Copyright 2025 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.
"""PyTorch Lfm2-VL model."""

from typing import Optional, Union

import torch
from torch import nn

from ...activations import ACT2FN
from ...cache_utils import Cache
from ...processing_utils import Unpack
from ...utils import TransformersKwargs, auto_docstring, can_return_tuple, logging
from ..llava.modeling_llava import (
    LlavaCausalLMOutputWithPast,
    LlavaForConditionalGeneration,
    LlavaModel,
    LlavaModelOutputWithPast,
    LlavaPreTrainedModel,
)
from .configuration_lfm2_vl import Lfm2VlConfig


logger = logging.get_logger(__name__)


class Lfm2VlMultiModalProjector(nn.Module):
    def __init__(self, config: Lfm2VlConfig):
        super().__init__()
        in_channels = config.vision_config.hidden_size * (config.downsample_factor**2)
        self.factor = config.downsample_factor
        self.layer_norm = nn.LayerNorm(in_channels)
        self.linear_1 = nn.Linear(
            in_channels,
            config.projector_hidden_size,
            bias=config.projector_bias,
        )
        self.act = ACT2FN[config.projector_hidden_act]
        self.linear_2 = nn.Linear(
            config.projector_hidden_size,
            config.text_config.hidden_size,
            bias=config.projector_bias,
        )

    def forward(self, image_features: torch.Tensor):
        image_features = self.pixel_unshuffle(image_features)
        image_features = self.layer_norm(image_features)
        hidden_states = self.linear_1(image_features)
        hidden_states = self.act(hidden_states)
        hidden_states = self.linear_2(hidden_states)
        return hidden_states

    def pixel_unshuffle(self, hidden_states: torch.Tensor):
        batch_size, width, height, channels = hidden_states.size()
        hidden_states = hidden_states.reshape(batch_size, width, height // self.factor, channels * self.factor)
        hidden_states = hidden_states.permute(0, 2, 1, 3)
        hidden_states = hidden_states.reshape(
            batch_size, height // self.factor, width // self.factor, channels * self.factor**2
        )
        hidden_states = hidden_states.permute(0, 2, 1, 3)
        return hidden_states


class Lfm2VlPreTrainedModel(LlavaPreTrainedModel):
    _can_compile_fullgraph = False


class Lfm2VlCausalLMOutputWithPast(LlavaCausalLMOutputWithPast):
    pass


class Lfm2VlModelOutputWithPast(LlavaModelOutputWithPast):
    pass


class Lfm2VlModel(LlavaModel):
    _checkpoint_conversion_mapping = {}

    def __init__(self, config: Lfm2VlConfig):
        super().__init__(config)

    def get_image_features(
        self,
        pixel_values: torch.FloatTensor,
        spatial_shapes: torch.Tensor,
        pixel_attention_mask: torch.Tensor,
        **kwargs,
    ) -> list[torch.Tensor]:
        """
        Obtains image last hidden states from the vision tower and apply multimodal projection.

        Args:
            pixel_values (`torch.FloatTensor]` of shape `(batch_size, channels, height, width)`):
               The tensors corresponding to the input images.
            spatial_shapes (`torch.Tensor` of shape `(batch_size, 2)`):
                The spatial shapes of the input images.
            pixel_attention_mask (`torch.Tensor` of shape `(batch_size, height, width)`):
                The pixel attention mask of the input images.
        Returns:
            image_features (`list[torch.Tensor]`): Image feature tensor of shape `(num_images, image_length, embed_dim)`).
        """
        image_outputs = self.vision_tower(
            pixel_values=pixel_values,
            spatial_shapes=spatial_shapes,
            pixel_attention_mask=pixel_attention_mask,
        ).last_hidden_state

        img_feature_lengths = pixel_attention_mask.sum(dim=1)
        image_features = []

        for img_idx in range(image_outputs.size(0)):
            feature = image_outputs[img_idx]
            # unpad the image representation
            feature = feature[: img_feature_lengths[img_idx], :].unsqueeze(0)

            # reshape to original height and width
            feature_org_h, feature_org_w = spatial_shapes[img_idx]
            feature = feature.reshape(1, feature_org_h, feature_org_w, -1)

            # project the image representation
            img_embedding = self.multi_modal_projector(feature)

            # flatten here to handle variable length in naflex
            img_embedding = img_embedding.reshape(-1, img_embedding.size(-1))
            image_features.append(img_embedding)

        return image_features

    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()
        special_image_mask = special_image_mask.unsqueeze(-1).expand_as(inputs_embeds).to(inputs_embeds.device)
        n_image_features = image_features.shape[0]
        if inputs_embeds[special_image_mask].numel() != image_features.numel():
            raise ValueError(
                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: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        pixel_values: Optional[torch.FloatTensor] = None,
        spatial_shapes: Optional[torch.Tensor] = None,
        pixel_attention_mask: Optional[torch.Tensor] = None,
        past_key_values: Optional[Cache] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        use_cache: Optional[bool] = None,
        cache_position: Optional[torch.LongTensor] = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> Union[tuple, Lfm2VlModelOutputWithPast]:
        r"""
        spatial_shapes (`torch.Tensor` of shape `(batch_size, 2)`, *optional*):
            The spatial shapes of the input images.
        pixel_attention_mask (`torch.Tensor` of shape `(batch_size, height, width)`, *optional*):
            The pixel attention mask of the input images.
        """

        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,
                spatial_shapes=spatial_shapes,
                pixel_attention_mask=pixel_attention_mask,
            )
            image_features = torch.cat(image_features, dim=0).to(inputs_embeds.device, inputs_embeds.dtype)
            special_image_mask = self.get_placeholder_mask(
                input_ids=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,
            cache_position=cache_position,
            **kwargs,
        )

        return Lfm2VlModelOutputWithPast(
            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,
        )


class Lfm2VlForConditionalGeneration(LlavaForConditionalGeneration):
    _checkpoint_conversion_mapping = {}

    def get_image_features(
        self,
        pixel_values: torch.FloatTensor,
        spatial_shapes: torch.Tensor,
        pixel_attention_mask: torch.Tensor,
        **kwargs,
    ):
        return self.model.get_image_features(
            pixel_values=pixel_values,
            spatial_shapes=spatial_shapes,
            pixel_attention_mask=pixel_attention_mask,
            **kwargs,
        )

    @can_return_tuple
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        pixel_values: Optional[torch.FloatTensor] = None,
        spatial_shapes: Optional[torch.Tensor] = None,
        pixel_attention_mask: Optional[torch.Tensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Cache] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        cache_position: Optional[torch.LongTensor] = None,
        logits_to_keep: Union[int, torch.Tensor] = 0,
        **kwargs: Unpack[TransformersKwargs],
    ) -> Union[tuple, Lfm2VlCausalLMOutputWithPast]:
        r"""
        pixel_values (`torch.FloatTensor` of shape `(batch_size, channels, height, width)`, *optional*):
            The input image tensors.
        spatial_shapes (`torch.Tensor` of shape `(batch_size, 2)`, *optional*):
            The spatial shapes of the input images.
        pixel_attention_mask (`torch.Tensor` of shape `(batch_size, height, width)`, *optional*):
            The pixel attention mask of the input images.
        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 requests
        >>> from transformers import AutoProcessor, AutoModelForImageTextToText
        >>> from transformers.image_utils import load_image

        >>> model = AutoModelForImageTextToText.from_pretrained(
        ...     "LiquidAI/LFM2-VL-1.6B",
        ... )
        >>> processor = AutoProcessor.from_pretrained(
        ...     "LiquidAI/LFM2-VL-1.6B",
        ... )

        >>> url = "https://www.ilankelman.org/stopsigns/australia.jpg"
        >>> image = load_image(url)

        >>> conversation = [
        ...     {
        ...         "role": "user",
        ...         "content": [
        ...             {"type": "image", "image": image},
        ...             {"type": "text", "text": "What is in this image?"},
        ...         ],
        ...     },
        ... ]

        >>> inputs = processor.apply_chat_template(
        ...     conversation,
        ...     add_generation_prompt=True,
        ...     tokenize=True,
        ...     return_dict=True,
        ...     return_tensors="pt"
        ... )

        >>> # Generate
        >>> outputs = model.generate(**inputs, max_new_tokens=45)
        >>> processor.batch_decode(outputs, skip_special_tokens=True)[0]
        'This image depicts a vibrant street scene in what appears to be a Chinatown or similar cultural area. The focal point is a large red stop sign with white lettering, mounted on a pole.'
        ```"""
        outputs = self.model(
            input_ids=input_ids,
            pixel_values=pixel_values,
            spatial_shapes=spatial_shapes,
            pixel_attention_mask=pixel_attention_mask,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            cache_position=cache_position,
            **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 Lfm2VlCausalLMOutputWithPast(
            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,
        )


__all__ = ["Lfm2VlForConditionalGeneration", "Lfm2VlPreTrainedModel", "Lfm2VlModel"]