ovis.py

# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project

# adapted from https://github.com/huggingface/transformers/blob/v4.39.3/src/transformers/models/ovis/modeling_ovis.py
# Copyright 2023 The vLLM team.
# Copyright 2023 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 Ovis model."""

import math
from collections.abc import Iterable, Mapping
from typing import Annotated, Literal

import torch
import torch.nn as nn
from torch import Tensor
from torch.nn.functional import gumbel_softmax, pad, softmax
from transformers import BatchFeature, PretrainedConfig

from vllm.config import VllmConfig
from vllm.config.multimodal import BaseDummyOptions
from vllm.inputs import MultiModalDataDict
from vllm.model_executor.layers.linear import ReplicatedLinear
from vllm.model_executor.layers.quantization import QuantizationConfig
from vllm.model_executor.models.aimv2 import AIMv2Model
from vllm.model_executor.models.siglip import SiglipVisionModel
from vllm.model_executor.models.utils import (
    AutoWeightsLoader,
    flatten_bn,
    init_vllm_registered_model,
    maybe_prefix,
)
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.multimodal.inputs import (
    MultiModalFieldConfig,
    MultiModalKwargsItems,
)
from vllm.multimodal.parse import ImageSize, MultiModalDataItems
from vllm.multimodal.processing import (
    BaseDummyInputsBuilder,
    BaseMultiModalProcessor,
    BaseProcessingInfo,
    PromptReplacement,
)
from vllm.renderers import TokenizeParams
from vllm.sequence import IntermediateTensors
from vllm.transformers_utils.processors.ovis import OvisProcessor
from vllm.utils.tensor_schema import TensorSchema, TensorShape

from .interfaces import MultiModalEmbeddings, SupportsMultiModal, SupportsPP

# Cannot find the following number from hf config.
IMAGE_TOKEN = "<image>"
IMAGE_INDICATOR_IDS = [-301, -302, -303, -304, -305]

IMAGE_PAD_TOKEN_MAP = {
    "gemma2": "<unused0>",
    "llama": "<|reserved_special_token_0|>",
    "qwen2": "<|image_pad|>",
}
IMAGE_PAD_TOKEN_ID_MAP = {
    "gemma2": 7,
    "llama": 128002,
    "qwen2": 151655,
}


def st_argmax(y_soft: torch.Tensor, dim: int):  # straight-through softmax
    index = y_soft.argmax(dim, keepdim=True)
    return torch.zeros_like(
        y_soft,
        memory_format=torch.legacy_contiguous_format,
    ).scatter_(dim, index, 1.0)


class VisualTokenizer(torch.nn.Module):
    def __init__(
        self,
        config: PretrainedConfig,
        quant_config: QuantizationConfig | None = None,
        prefix: str = "",
    ):
        super().__init__()
        self.config = config
        self.backbone = self._init_backbone(
            config=config,
            quant_config=quant_config,
            prefix=f"{prefix}.backbone",
        )
        # reserved tokens for IMAGE_INDICATORS
        head_dim = config.vocab_size - len(IMAGE_INDICATOR_IDS)
        self.head = torch.nn.Sequential(
            ReplicatedLinear(
                config.backbone_config.hidden_size
                * config.hidden_stride
                * config.hidden_stride,
                head_dim,
                bias=False,
                return_bias=False,
            ),
            torch.nn.LayerNorm(head_dim),
        )

    def _init_backbone(
        self,
        config: PretrainedConfig,
        quant_config: QuantizationConfig | None = None,
        prefix: str = "",
    ) -> nn.Module:
        model_type = config.backbone_config.model_type
        if model_type == "aimv2":
            # No post rms_norm in Ovis2's AIMv2 ViT.
            return AIMv2Model(
                config=config.backbone_config,
                quant_config=quant_config,
                require_post_norm=False,
                prefix=prefix,
            )
        elif model_type == "siglip_vision_model":
            return SiglipVisionModel(
                config=config.backbone_config,
                quant_config=quant_config,
                prefix=prefix,
            )
        raise ValueError(f"Unsupported visual tokenizer model_type: {model_type}")

    @property
    def dtype(self) -> torch.dtype:
        return next(self.head.parameters()).dtype

    @property
    def device(self) -> torch.device:
        return next(self.head.parameters()).device

    def tokenize(self, logits: torch.Tensor) -> torch.Tensor:
        if self.config.tokenize_function == "softmax":
            tokens = softmax(logits, dim=-1)
        elif self.config.tokenize_function == "gumbel_argmax":
            tokens = gumbel_softmax(logits, tau=self.config.tau, hard=True)
        elif self.config.tokenize_function == "st_argmax":
            tokens = st_argmax(logits, dim=-1)
        else:
            raise ValueError(
                "Invalid `max_type`, expected softmax or gumbel_argmax "
                f"or st_argmax, but got {self.config.tokenize_function}"
            )
        return tokens

    def encode(self, pixel_values: torch.Tensor) -> torch.Tensor:
        features = self.backbone(pixel_values)
        if self.config.drop_cls_token:
            features = features[:, 1:, :]

        # merge number of `hidden_stride * hidden_stride` hidden states together
        # to reduce token sequence length
        # e.g., for hidden_stride=2, this leads to a token length reduction:
        # 1024 -> 256 for aimv2
        if self.config.hidden_stride > 1:
            # this `d` maybe different from the above `d`
            n, L, d = features.shape
            sqrt_l = int(L**0.5)
            assert sqrt_l**2 == L, (
                "The token sequence length should be a perfect square."
            )
            features = features.reshape(n, sqrt_l, sqrt_l, d)
            pl = (
                self.config.hidden_stride - (sqrt_l % self.config.hidden_stride)
            ) % self.config.hidden_stride
            features = pad(features, (0, 0, 0, pl, 0, pl), "constant", 0)
            sqrt_l += pl
            features = features.reshape(
                n,
                sqrt_l // self.config.hidden_stride,
                self.config.hidden_stride,
                sqrt_l // self.config.hidden_stride,
                self.config.hidden_stride,
                d,
            )
            # [n, sqrt_l/hs, sqrt_l/hs, hs, hs, d]
            features = features.permute(0, 1, 3, 2, 4, 5)
            # [n, sqrt_l/hs, sqrt_l/hs, hs*hs*d]
            features = features.flatten(3)
            # [n, sqrt_l/hs*sqrt_l/hs, hs*hs*d]
            features = features.reshape(
                n, -1, self.config.hidden_stride * self.config.hidden_stride * d
            )

        return features

    def forward(self, pixel_values: torch.Tensor) -> torch.Tensor:
        """[BatchSize, ImageShape] -> [BatchSize, Token, VocabSize]"""
        features = self.encode(pixel_values)
        logits = self.head(features)
        tokens = self.tokenize(logits)
        # tokens' shape is [BatchSize, #Token, VocabSize-5], so padding with
        # [BatchSize, #Token, 5], after which, tokens' shape should become
        # [BatchSize, #Token, VocabSize]
        tokens = torch.nn.functional.pad(
            tokens,
            (0, len(IMAGE_INDICATOR_IDS)),
            mode="constant",
            value=0,
        )
        return tokens


class OvisImagePatchInputs(TensorSchema):
    """
    Dimensions:
        - bnp: Batch size * number of images * number of patches
        - h: Height of each patch
        - w: Width of each patch
        - patch_indicators: Batch size * (number of patches + 1)
        - bn: Batch size * number of images
    """

    type: Literal["image_patches"]
    flat_data: Annotated[torch.Tensor, TensorShape("bnp", 3, "h", "w")]
    indicator_tokens: Annotated[torch.Tensor, TensorShape("patch_indicators")]
    patches_per_image: Annotated[list[int], TensorShape("bn")]
    # This is used to restore the first two dimensions of `flat_data`.


class VisualEmbedding(torch.nn.Embedding):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    def forward(self, visual_tokens: Tensor) -> Tensor:
        if visual_tokens.dtype in [
            torch.int8,
            torch.int16,
            torch.int32,
            torch.int64,
            torch.long,
        ]:
            return super().forward(visual_tokens)
        return torch.matmul(visual_tokens, self.weight)

    @property
    def device(self):
        return self.weight.device

    @property
    def dtype(self):
        return self.weight.dtype


class OvisProcessingInfo(BaseProcessingInfo):
    def get_hf_processor(self, **kwargs: object):
        return self.ctx.get_hf_processor(
            OvisProcessor,
            image_pad_token=self.get_image_pad_token(),
            image_segment_len=self.get_image_segment_len(),
            **kwargs,
        )

    def get_default_tok_params(self) -> TokenizeParams:
        return super().get_default_tok_params().with_kwargs(add_special_tokens=False)

    def get_image_segment_len(self) -> int:
        visual_tokenizer_config = self.get_hf_config().visual_tokenizer_config
        image_size = visual_tokenizer_config.backbone_config.image_size
        patch_size = visual_tokenizer_config.backbone_config.patch_size
        hidden_stride = visual_tokenizer_config.hidden_stride
        patch_grid_length = math.ceil(image_size / patch_size)
        assert patch_grid_length % hidden_stride == 0, (
            f"patch_grid_length {patch_grid_length} is not divisible by "
            f"hidden_stride {hidden_stride}"
        )
        # minus 1 for presented image token
        return (patch_grid_length // hidden_stride) ** 2 - 1

    def get_image_pad_token(self) -> str:
        hf_text_config = self.get_hf_config().get_text_config()
        text_model_type = hf_text_config.model_type
        return IMAGE_PAD_TOKEN_MAP.get(text_model_type)

    def get_supported_mm_limits(self) -> Mapping[str, int | None]:
        return {"image": None}

    def get_image_size_with_most_features(self) -> ImageSize:
        height, width = self.get_hf_processor().get_image_size()
        hs = self.get_hf_config().visual_tokenizer_config.hidden_stride
        # NOTE(Isotr0py): 9 is `max_partition` hardcoded in original code
        # https://huggingface.co/AIDC-AI/Ovis2-1B/blob/main/modeling_ovis.py#L96
        return ImageSize(width=width * hs * 9, height=height * hs * 9)


class OvisDummyInputsBuilder(BaseDummyInputsBuilder[OvisProcessingInfo]):
    def get_dummy_text(self, mm_counts: Mapping[str, int]) -> str:
        num_images = mm_counts.get("image", 0)
        return IMAGE_TOKEN * num_images

    def get_dummy_mm_data(
        self,
        seq_len: int,
        mm_counts: Mapping[str, int],
        mm_options: Mapping[str, BaseDummyOptions],
    ) -> MultiModalDataDict:
        num_images = mm_counts.get("image", 0)

        target_width, target_height = self.info.get_image_size_with_most_features()

        image_overrides = mm_options.get("image")

        mm_data = {
            "image": self._get_dummy_images(
                width=target_width,
                height=target_height,
                num_images=num_images,
                overrides=image_overrides,
            ),
        }
        return mm_data


class OvisMultiModalProcessor(BaseMultiModalProcessor[OvisProcessingInfo]):
    def image_indicators_to_visual_tokens(
        self,
        image_indicators: list[int],
    ) -> list[int]:
        """
        Filter image indicators placeholders and convert them to corresponding
        tokens in visual tokenizer.
        For example, [-301, -300, -302, -300, -303, -300, -304, -300, -305]
        should return [vocab_size-1, vocab_size-2, ..., vocab_size-5]
        """
        hf_config = self.info.get_hf_config()
        vte_vocab_size = hf_config.visual_tokenizer_config.vocab_size
        # -300 is image_atom token, filter them out
        return [vte_vocab_size + x + 300 for x in image_indicators if x < -300]

    def _call_hf_processor(
        self,
        prompt: str,
        mm_data: Mapping[str, object],
        mm_kwargs: Mapping[str, object],
        tok_kwargs: Mapping[str, object],
    ) -> BatchFeature:
        if not mm_data:
            # Avoid warning from HF logger for text-only input
            tokenizer = self.info.get_tokenizer()
            prompt_ids = tokenizer.encode(prompt, add_special_tokens=False)
            return BatchFeature(dict(input_ids=[prompt_ids]), tensor_type="pt")

        processed_outputs = super()._call_hf_processor(
            prompt=prompt,
            mm_data=mm_data,
            mm_kwargs=mm_kwargs,
            tok_kwargs=tok_kwargs,
        )

        hf_processor = self.info.get_hf_processor()
        image_indicators = [
            hf_processor.construct_image_indicators(grid)
            for grid in processed_outputs["grids"]
        ]
        indicator_tokens = [
            self.image_indicators_to_visual_tokens(indicator)
            for indicator in image_indicators
        ]
        processed_outputs["indicator_tokens"] = torch.tensor(indicator_tokens)
        return processed_outputs

    def _apply_hf_processor_tokens_only(
        self,
        prompt_tokens: list[int],
    ) -> list[int]:
        return prompt_tokens

    def _get_mm_fields_config(
        self,
        hf_inputs: BatchFeature,
        hf_processor_mm_kwargs: Mapping[str, object],
    ) -> Mapping[str, MultiModalFieldConfig]:
        return dict(
            pixel_values=MultiModalFieldConfig.batched("image"),
            grids=MultiModalFieldConfig.batched("image"),
            indicator_tokens=MultiModalFieldConfig.batched("image"),
        )

    def _get_prompt_updates(
        self,
        mm_items: MultiModalDataItems,
        hf_processor_mm_kwargs: Mapping[str, object],
        out_mm_kwargs: MultiModalKwargsItems,
    ) -> list[PromptReplacement]:
        def get_replacement_ovis(item_idx: int):
            out_item = out_mm_kwargs["image"][item_idx]
            grid = out_item["grids"].data

            hf_processor = self.info.get_hf_processor()
            return hf_processor.construct_image_placeholders(grid)

        return [
            PromptReplacement(
                modality="image",
                target=IMAGE_TOKEN,
                replacement=get_replacement_ovis,
            ),
        ]


@MULTIMODAL_REGISTRY.register_processor(
    OvisMultiModalProcessor,
    info=OvisProcessingInfo,
    dummy_inputs=OvisDummyInputsBuilder,
)
class Ovis(nn.Module, SupportsMultiModal, SupportsPP):
    @classmethod
    def get_placeholder_str(cls, modality: str, i: int) -> str | None:
        if modality.startswith("image"):
            return IMAGE_TOKEN

        raise ValueError("Only image modality is supported")

    def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
        super().__init__()
        config = vllm_config.model_config.hf_config
        quant_config = vllm_config.quant_config

        self.config: PretrainedConfig = config

        with self._mark_language_model(vllm_config):
            self.llm = init_vllm_registered_model(
                vllm_config=vllm_config.with_hf_config(config.get_text_config()),
                prefix=maybe_prefix(prefix, "llm"),
            )

        with self._mark_tower_model(vllm_config, "image"):
            self.visual_tokenizer = VisualTokenizer(
                config=config.visual_tokenizer_config,
                quant_config=quant_config,
                prefix=f"{prefix}.visual_tokenizer",
            )
            self.vte = VisualEmbedding(
                self.config.visual_tokenizer_config.vocab_size, self.config.hidden_size
            )

        text_model_type = self.config.get_text_config().model_type
        self.image_pad_token_id = IMAGE_PAD_TOKEN_ID_MAP[text_model_type]

        self.make_empty_intermediate_tensors = (
            self.get_language_model().make_empty_intermediate_tensors
        )

    def _parse_and_validate_image_input(
        self, **kwargs: object
    ) -> OvisImagePatchInputs | None:
        pixel_values = kwargs.pop("pixel_values", None)
        indicator_tokens = kwargs.pop("indicator_tokens", None)

        if pixel_values is None and indicator_tokens is None:
            return None

        if pixel_values is not None and indicator_tokens is not None:
            if not isinstance(pixel_values, (torch.Tensor, list)):
                raise ValueError(
                    f"Incorrect type of pixel values. Got type: {type(pixel_values)}"
                )

            if not isinstance(indicator_tokens, (torch.Tensor, list)):
                raise ValueError(
                    "Incorrect type of indicator_tokens. "
                    f"Got type: {type(pixel_values)}"
                )

            return OvisImagePatchInputs(
                type="image_patches",
                flat_data=flatten_bn(pixel_values, concat=True),
                patches_per_image=[x.shape[0] for x in pixel_values],
                indicator_tokens=flatten_bn(indicator_tokens, concat=True),
            )

        raise AssertionError("This line should be unreachable.")

    def _process_image_input(
        self, image_input: OvisImagePatchInputs
    ) -> MultiModalEmbeddings:
        image_patches_flat = image_input["flat_data"]
        patches_per_image = image_input["patches_per_image"]
        indicator_tokens = image_input["indicator_tokens"]

        indicator_per_image = list(
            map(lambda x: x + 1 if x > 1 else x + 2, patches_per_image)
        )

        target_dtype = self.visual_tokenizer.dtype
        visual_tokens = self.visual_tokenizer(image_patches_flat.to(target_dtype))
        visual_embeds = self.vte(visual_tokens)  # 1:1 numeric eq.

        indicator_embeds = self.vte(indicator_tokens)
        indicator_embeds_per_image = indicator_embeds.split(indicator_per_image)

        visual_embeds_per_image = visual_embeds.split(patches_per_image, dim=0)
        vision_embeddings = []
        for indicator, visual in zip(
            indicator_embeds_per_image, visual_embeds_per_image
        ):
            vision_embeddings_per_image = []
            for i in range(visual.shape[0]):
                vision_embeddings_per_image.append(
                    torch.cat([indicator[i : i + 1], visual[i]], dim=0)
                )
            vision_embeddings_per_image.append(indicator[i + 1 :])
            vision_embeddings.append(torch.cat(vision_embeddings_per_image, dim=0))

        return tuple(vision_embeddings)

    def embed_multimodal(self, **kwargs: object) -> MultiModalEmbeddings:
        image_input = self._parse_and_validate_image_input(**kwargs)
        if image_input is None:
            return []

        image_features = self._process_image_input(image_input)

        return image_features

    def forward(
        self,
        input_ids: torch.Tensor | None,
        positions: torch.Tensor,
        intermediate_tensors: IntermediateTensors | None = None,
        inputs_embeds: torch.Tensor | None = None,
        **kwargs: object,
    ) -> torch.Tensor | IntermediateTensors:
        if intermediate_tensors is not None:
            inputs_embeds = None

        # up until here we have an inputs_embeds 100% numerical identity
        # between the OG HF Transformers implementation and ours
        hidden_states = self.llm(
            input_ids=input_ids,
            positions=positions,
            intermediate_tensors=intermediate_tensors,
            inputs_embeds=inputs_embeds,
        )
        return hidden_states

    def compute_logits(
        self,
        hidden_states: torch.Tensor,
    ) -> torch.Tensor | None:
        return self.llm.compute_logits(hidden_states)

    def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
        loader = AutoWeightsLoader(self)
        return loader.load_weights(weights)