llava.py

from typing import Iterable, List, Literal, Optional, Tuple, TypedDict

import torch
import torch.nn as nn
from transformers import CLIPVisionConfig, LlavaConfig

from vllm.attention import AttentionMetadata
from vllm.config import CacheConfig, MultiModalConfig
from vllm.inputs import INPUT_REGISTRY, InputContext, LLMInputs
from vllm.model_executor.layers.activation import get_act_fn
from vllm.model_executor.layers.logits_processor import LogitsProcessor
from vllm.model_executor.layers.quantization.base_config import (
    QuantizationConfig)
from vllm.model_executor.layers.sampler import Sampler
from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.model_executor.models.clip import CLIPVisionModel
from vllm.model_executor.models.llama import LlamaModel
from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.sequence import IntermediateTensors, SamplerOutput

from .clip import (dummy_image_for_clip, dummy_seq_data_for_clip,
                   get_max_clip_image_tokens, input_processor_for_clip)
from .interfaces import SupportsVision
from .utils import merge_vision_embeddings

_KEYS_TO_MODIFY_MAPPING = {
    "language_model.lm_head": "lm_head",
    "language_model.model": "language_model",
}


# TODO(xwjiang): Run benchmark and decide if TP.
class LlavaMultiModalProjector(nn.Module):

    def __init__(self, vision_hidden_size: int, text_hidden_size: int,
                 projector_hidden_act: str):
        super().__init__()

        self.linear_1 = nn.Linear(vision_hidden_size,
                                  text_hidden_size,
                                  bias=True)
        self.act = get_act_fn(projector_hidden_act)
        self.linear_2 = nn.Linear(text_hidden_size,
                                  text_hidden_size,
                                  bias=True)

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


class LlavaImagePixelInputs(TypedDict):
    type: Literal["pixel_values"]
    data: torch.Tensor
    """Shape: `(batch_size, num_channels, height, width)`"""


LlavaImageInputs = LlavaImagePixelInputs


def get_max_llava_image_tokens(ctx: InputContext):
    hf_config = ctx.get_hf_config(LlavaConfig)
    vision_config = hf_config.vision_config

    if isinstance(vision_config, CLIPVisionConfig):
        return get_max_clip_image_tokens(vision_config)

    msg = f"Unsupported vision config: {type(vision_config)}"
    raise NotImplementedError(msg)


def dummy_data_for_llava(ctx: InputContext, seq_len: int):
    hf_config = ctx.get_hf_config(LlavaConfig)
    vision_config = hf_config.vision_config

    if isinstance(vision_config, CLIPVisionConfig):
        seq_data = dummy_seq_data_for_clip(
            vision_config,
            seq_len,
            image_token_id=hf_config.image_token_index,
        )

        mm_data = dummy_image_for_clip(vision_config)
        return seq_data, mm_data

    msg = f"Unsupported vision config: {type(vision_config)}"
    raise NotImplementedError(msg)


def input_processor_for_llava(ctx: InputContext, llm_inputs: LLMInputs):
    multi_modal_data = llm_inputs.get("multi_modal_data")
    if multi_modal_data is None or "image" not in multi_modal_data:
        return llm_inputs

    model_config = ctx.model_config
    hf_config = ctx.get_hf_config(LlavaConfig)
    vision_config = hf_config.vision_config

    if isinstance(vision_config, CLIPVisionConfig):
        return input_processor_for_clip(
            model_config,
            vision_config,
            llm_inputs,
            image_token_id=hf_config.image_token_index,
        )

    msg = f"Unsupported vision config: {type(vision_config)}"
    raise NotImplementedError(msg)


@MULTIMODAL_REGISTRY.register_image_input_mapper()
@MULTIMODAL_REGISTRY.register_max_image_tokens(get_max_llava_image_tokens)
@INPUT_REGISTRY.register_dummy_data(dummy_data_for_llava)
@INPUT_REGISTRY.register_input_processor(input_processor_for_llava)
class LlavaForConditionalGeneration(nn.Module, SupportsVision):

    def __init__(self,
                 config: LlavaConfig,
                 multimodal_config: MultiModalConfig,
                 cache_config: Optional[CacheConfig] = None,
                 quant_config: Optional[QuantizationConfig] = None) -> None:
        super().__init__()

        self.config = config
        self.multimodal_config = multimodal_config

        # Initialize the vision tower only up to the required feature layer
        vision_feature_layer = config.vision_feature_layer
        if vision_feature_layer < 0:
            num_hidden_layers = config.vision_config.num_hidden_layers \
                + vision_feature_layer + 1
        else:
            num_hidden_layers = vision_feature_layer + 1

        # TODO: Optionally initializes this for supporting embeddings.
        self.vision_tower = CLIPVisionModel(
            config.vision_config, num_hidden_layers_override=num_hidden_layers)
        self.multi_modal_projector = LlavaMultiModalProjector(
            vision_hidden_size=config.vision_config.hidden_size,
            text_hidden_size=config.text_config.hidden_size,
            projector_hidden_act=config.projector_hidden_act)

        self.quant_config = quant_config
        self.language_model = LlamaModel(config.text_config, cache_config,
                                         quant_config)
        self.unpadded_vocab_size = config.text_config.vocab_size
        self.lm_head = ParallelLMHead(
            self.unpadded_vocab_size,
            config.text_config.hidden_size,
            org_num_embeddings=self.language_model.org_vocab_size,
            quant_config=quant_config)
        logit_scale = getattr(config, "logit_scale", 1.0)
        self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                config.vocab_size, logit_scale)
        self.sampler = Sampler()

    def _validate_pixel_values(self, data: torch.Tensor) -> torch.Tensor:
        h = w = self.config.vision_config.image_size
        expected_dims = (3, h, w)
        actual_dims = tuple(data.shape[1:])

        if actual_dims != expected_dims:
            expected_expr = ("batch_size", *map(str, expected_dims))
            raise ValueError(
                f"The expected shape of pixel values is {expected_expr}. "
                f"You supplied {tuple(data.shape)}.")

        return data

    def _parse_and_validate_image_input(
            self, **kwargs: object) -> Optional[LlavaImageInputs]:
        pixel_values = kwargs.pop("pixel_values", None)

        if pixel_values is None:
            return None

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

        return LlavaImagePixelInputs(
            type="pixel_values",
            data=self._validate_pixel_values(pixel_values),
        )

    def _select_image_features(self, image_features: torch.Tensor, *,
                               strategy: str) -> torch.Tensor:
        # Copied from https://github.com/huggingface/transformers/blob/39c3c0a72af6fbda5614dde02ff236069bb79827/src/transformers/models/llava/modeling_llava.py#L421  # noqa
        if strategy == "default":
            return image_features[:, 1:]
        elif strategy == "full":
            return image_features

        raise ValueError(f"Unexpected select feature strategy: {strategy}")

    def _image_pixels_to_features(self, vision_tower: CLIPVisionModel,
                                  pixel_values: torch.Tensor) -> torch.Tensor:

        # NOTE: we skip the step to select the vision feature layer since
        # this is already done inside the vision tower
        image_features = vision_tower(pixel_values)

        return self._select_image_features(
            image_features,
            strategy=self.config.vision_feature_select_strategy,
        )

    def _process_image_pixels(self,
                              inputs: LlavaImagePixelInputs) -> torch.Tensor:
        assert self.vision_tower is not None

        pixel_values = inputs["data"]

        return self._image_pixels_to_features(self.vision_tower, pixel_values)

    def _process_image_input(self,
                             image_input: LlavaImageInputs) -> torch.Tensor:
        assert self.vision_tower is not None
        image_features = self._process_image_pixels(image_input)
        return self.multi_modal_projector(image_features)

    def forward(
        self,
        input_ids: torch.Tensor,
        positions: torch.Tensor,
        kv_caches: List[torch.Tensor],
        attn_metadata: AttentionMetadata,
        intermediate_tensors: Optional[IntermediateTensors] = None,
        **kwargs: object,
    ) -> SamplerOutput:
        """Run forward pass for LLaVA-1.5.

        One key thing to understand is the `input_ids` already accounts for the
        positions of the to-be-inserted image embeddings.

        Concretely, consider a text prompt:
        `"USER: <image>\\nWhat's the content of the image?\\nASSISTANT:"`.

        Tokenizer outputs:
        `[1, 3148, 1001, 29901, 29871, 32000, 29871, 13, 5618, 29915, 29879,
        278, 2793, 310, 278, 1967, 29973, 13, 22933, 9047, 13566, 29901]`.

        To reserve space in KV cache, we have to insert placeholder tokens
        before they are inputted to the model, so the input processor prepends 
        additional image tokens (denoted as `32000`), resulting in:
        `[1, 3148, 1001, 29901, 29871, 32000, ..., 32000, 29871, 13, 5618,
        29915, 29879, 278, 2793, 310, 278, 1967, 29973, 13, 22933, 9047, 13566,
        29901]`.

        We insert 575 tokens so that including the original image token in the
        input, there are a total of 576 (24 * 24) image tokens, which
        corresponds to the number of image tokens inputted to the language
        model, i.e. the number of image tokens outputted by the visual encoder.

        This way, the `positions` and `attn_metadata` are consistent
        with the `input_ids`.

        Args:
            input_ids: Flattened (concatenated) input_ids corresponding to a
                batch.
            pixel_values: The pixels in each input image.
        
        See also:
            :class:`LlavaImageInputs`
        """
        image_input = self._parse_and_validate_image_input(**kwargs)

        if image_input is not None:
            vision_embeddings = self._process_image_input(image_input)
            inputs_embeds = self.language_model.get_input_embeddings(input_ids)

            inputs_embeds = merge_vision_embeddings(
                input_ids, inputs_embeds, vision_embeddings,
                self.config.image_token_index)

            input_ids = None
        else:
            inputs_embeds = None

        hidden_states = self.language_model(input_ids,
                                            positions,
                                            kv_caches,
                                            attn_metadata,
                                            None,
                                            inputs_embeds=inputs_embeds)

        return hidden_states

    def compute_logits(self, hidden_states: torch.Tensor,
                       sampling_metadata: SamplingMetadata) -> torch.Tensor:
        logits = self.logits_processor(self.lm_head, hidden_states,
                                       sampling_metadata)
        return logits

    def sample(
        self,
        logits: torch.Tensor,
        sampling_metadata: SamplingMetadata,
    ) -> Optional[SamplerOutput]:
        next_tokens = self.sampler(logits, sampling_metadata)
        return next_tokens

    def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
        # only doing this for language model part for now.
        stacked_params_mapping = [
            # (param_name, shard_name, shard_id)
            ("qkv_proj", "q_proj", "q"),
            ("qkv_proj", "k_proj", "k"),
            ("qkv_proj", "v_proj", "v"),
            ("gate_up_proj", "gate_proj", 0),
            ("gate_up_proj", "up_proj", 1),
        ]
        params_dict = dict(self.named_parameters())
        for name, loaded_weight in weights:
            if "rotary_emb.inv_freq" in name:
                continue
            # post_layernorm is not needed in CLIPVisionModel
            if "vision_model.post_layernorm" in name:
                continue
            for key_to_modify, new_key in _KEYS_TO_MODIFY_MAPPING.items():
                if key_to_modify in name:
                    name = name.replace(key_to_modify, new_key)
            use_default_weight_loading = False
            if "vision" in name:
                if self.vision_tower is not None:
                    # We only do sharding for language model and
                    # not vision model for now.
                    use_default_weight_loading = True
            else:
                for (param_name, weight_name,
                     shard_id) in stacked_params_mapping:
                    if weight_name not in name:
                        continue
                    param = params_dict[name.replace(weight_name, param_name)]
                    weight_loader = param.weight_loader
                    weight_loader(param, loaded_weight, shard_id)
                    break
                else:
                    use_default_weight_loading = True
            if use_default_weight_loading and name in params_dict:
                param = params_dict[name]
                weight_loader = getattr(param, "weight_loader",
                                        default_weight_loader)
                weight_loader(param, loaded_weight)