[Model][HunyuanVL] Use native transformers processor and adapt to transformers 5.13 (#47872)
Co-authored-by: manayang <manayang@tencent.com>
This commit is contained in:
@@ -936,7 +936,7 @@ _MULTIMODAL_EXAMPLE_MODELS = {
|
|||||||
"HunYuanVLForConditionalGeneration": _HfExamplesInfo(
|
"HunYuanVLForConditionalGeneration": _HfExamplesInfo(
|
||||||
"tencent/HunyuanOCR",
|
"tencent/HunyuanOCR",
|
||||||
hf_overrides={"num_experts": 0},
|
hf_overrides={"num_experts": 0},
|
||||||
is_available_online=False,
|
min_transformers_version="5.13",
|
||||||
),
|
),
|
||||||
"Idefics3ForConditionalGeneration": _HfExamplesInfo(
|
"Idefics3ForConditionalGeneration": _HfExamplesInfo(
|
||||||
"HuggingFaceM4/Idefics3-8B-Llama3",
|
"HuggingFaceM4/Idefics3-8B-Llama3",
|
||||||
|
|||||||
@@ -31,7 +31,11 @@ from typing import Annotated, Any, Literal, TypeAlias
|
|||||||
import torch
|
import torch
|
||||||
import torch.nn as nn
|
import torch.nn as nn
|
||||||
import torch.nn.functional as F
|
import torch.nn.functional as F
|
||||||
from transformers import BatchFeature
|
from transformers import BatchFeature, HunYuanVLProcessor
|
||||||
|
from transformers.models.hunyuan_vl.image_processing_hunyuan_vl import (
|
||||||
|
HunYuanVLImageProcessor,
|
||||||
|
smart_resize,
|
||||||
|
)
|
||||||
|
|
||||||
from vllm.config import VllmConfig
|
from vllm.config import VllmConfig
|
||||||
from vllm.config.multimodal import BaseDummyOptions
|
from vllm.config.multimodal import BaseDummyOptions
|
||||||
@@ -75,11 +79,6 @@ from vllm.transformers_utils.configs.hunyuan_vl import (
|
|||||||
HunYuanVLConfig,
|
HunYuanVLConfig,
|
||||||
HunYuanVLVisionConfig,
|
HunYuanVLVisionConfig,
|
||||||
)
|
)
|
||||||
from vllm.transformers_utils.processors.hunyuan_vl import HunYuanVLProcessor
|
|
||||||
from vllm.transformers_utils.processors.hunyuan_vl_image import (
|
|
||||||
HunYuanVLImageProcessor,
|
|
||||||
smart_resize,
|
|
||||||
)
|
|
||||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||||
|
|
||||||
from .interfaces import (
|
from .interfaces import (
|
||||||
@@ -574,9 +573,12 @@ class HunYuanVLProcessingInfo(BaseProcessingInfo):
|
|||||||
self,
|
self,
|
||||||
**kwargs: object,
|
**kwargs: object,
|
||||||
) -> HunYuanVLProcessor:
|
) -> HunYuanVLProcessor:
|
||||||
|
# transformers>=5.13 replaced `use_fast` with `backend`; pin the
|
||||||
|
# PIL backend to match the released HunyuanOCR checkpoint packing.
|
||||||
|
kwargs.pop("use_fast", None)
|
||||||
|
kwargs.setdefault("backend", "pil")
|
||||||
return self.ctx.get_hf_processor(
|
return self.ctx.get_hf_processor(
|
||||||
HunYuanVLProcessor,
|
HunYuanVLProcessor,
|
||||||
use_fast=kwargs.pop("use_fast", True),
|
|
||||||
**kwargs,
|
**kwargs,
|
||||||
)
|
)
|
||||||
|
|
||||||
@@ -724,8 +726,18 @@ class HunYuanVLMultiModalProcessor(BaseMultiModalProcessor[HunYuanVLProcessingIn
|
|||||||
mm_kwargs: Mapping[str, object],
|
mm_kwargs: Mapping[str, object],
|
||||||
tok_kwargs: Mapping[str, object],
|
tok_kwargs: Mapping[str, object],
|
||||||
) -> BatchFeature:
|
) -> BatchFeature:
|
||||||
|
hf_processor = self.info.get_hf_processor(**mm_kwargs)
|
||||||
|
# HunYuanVLProcessor requires image placeholders wrapped with start/end tokens.
|
||||||
|
if mm_data.get("images") is not None and prompt:
|
||||||
|
img_tok = hf_processor.image_token
|
||||||
|
wrapped = (
|
||||||
|
f"{hf_processor.image_start_token}{img_tok}"
|
||||||
|
f"{hf_processor.image_end_token}"
|
||||||
|
)
|
||||||
|
if img_tok in prompt and wrapped not in prompt:
|
||||||
|
prompt = prompt.replace(img_tok, wrapped)
|
||||||
return self.info.ctx.call_hf_processor(
|
return self.info.ctx.call_hf_processor(
|
||||||
self.info.get_hf_processor(**mm_kwargs),
|
hf_processor,
|
||||||
dict(text=prompt, **mm_data),
|
dict(text=prompt, **mm_data),
|
||||||
dict(**mm_kwargs, **tok_kwargs),
|
dict(**mm_kwargs, **tok_kwargs),
|
||||||
)
|
)
|
||||||
|
|||||||
@@ -1,226 +0,0 @@
|
|||||||
# SPDX-License-Identifier: Apache-2.0
|
|
||||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
|
||||||
|
|
||||||
# adapted from https://github.com/ManaEstras/transformers/blob/v4.57.1.hyvl/src/transformers/models/hunyuan_vl/processing_hunyuan_vl.py
|
|
||||||
|
|
||||||
import numpy as np
|
|
||||||
import torch
|
|
||||||
from transformers.feature_extraction_utils import BatchFeature
|
|
||||||
from transformers.image_utils import ImageInput
|
|
||||||
from transformers.processing_utils import ProcessorMixin
|
|
||||||
from transformers.tokenization_utils_base import PreTokenizedInput, TextInput
|
|
||||||
from transformers.video_utils import VideoInput
|
|
||||||
|
|
||||||
|
|
||||||
class HunYuanVLProcessor(ProcessorMixin):
|
|
||||||
attributes = ["image_processor", "tokenizer"]
|
|
||||||
valid_kwargs = ["chat_template"]
|
|
||||||
image_processor_class = "AutoImageProcessor"
|
|
||||||
tokenizer_class = "AutoTokenizer" # ("AutoTokenizer", None)
|
|
||||||
|
|
||||||
def __init__(
|
|
||||||
self,
|
|
||||||
image_processor=None,
|
|
||||||
tokenizer=None,
|
|
||||||
chat_template=None,
|
|
||||||
**kwargs,
|
|
||||||
):
|
|
||||||
# TODO Fix the init
|
|
||||||
self.tokenizer = tokenizer
|
|
||||||
self.image_token_id = 120120 # self.tokenizer.image_token_id
|
|
||||||
self.image_token = self.tokenizer.convert_ids_to_tokens(self.image_token_id)
|
|
||||||
self.im_start_token_id = 120118 # self.tokenizer.im_start_id
|
|
||||||
self.im_start_token = self.tokenizer.convert_ids_to_tokens(
|
|
||||||
self.im_start_token_id
|
|
||||||
)
|
|
||||||
self.im_end_token_id = 120119 # self.tokenizer.im_end_id
|
|
||||||
self.im_end_token = self.tokenizer.convert_ids_to_tokens(self.im_end_token_id)
|
|
||||||
self.placeholder_token = self.tokenizer.convert_ids_to_tokens(
|
|
||||||
self.tokenizer.vocab_size - 1
|
|
||||||
)
|
|
||||||
self.pad_id = 120002 # self.tokenizer.pad_token_id
|
|
||||||
|
|
||||||
super().__init__(image_processor, tokenizer, chat_template=chat_template)
|
|
||||||
|
|
||||||
def __call__(
|
|
||||||
self,
|
|
||||||
images: ImageInput = None,
|
|
||||||
text: TextInput
|
|
||||||
| PreTokenizedInput
|
|
||||||
| list[TextInput]
|
|
||||||
| list[PreTokenizedInput] = None,
|
|
||||||
videos: VideoInput = None,
|
|
||||||
**kwargs,
|
|
||||||
) -> BatchFeature:
|
|
||||||
image_inputs = {}
|
|
||||||
if images is not None:
|
|
||||||
image_inputs = self.image_processor(images=images)
|
|
||||||
image_grid_thw = image_inputs["image_grid_thw"]
|
|
||||||
|
|
||||||
if not isinstance(text, list):
|
|
||||||
text = [text]
|
|
||||||
|
|
||||||
text = text.copy() # below lines change text in-place
|
|
||||||
|
|
||||||
image_tokens_cumsum = [0]
|
|
||||||
if images is not None:
|
|
||||||
index = 0
|
|
||||||
for i in range(len(text)):
|
|
||||||
while self.image_token in text[i]:
|
|
||||||
grid_h, grid_w = image_grid_thw[index][-2:]
|
|
||||||
patch_h = grid_h // self.image_processor.merge_size
|
|
||||||
patch_w = grid_w // self.image_processor.merge_size
|
|
||||||
num_image_tokens = patch_h * (patch_w + 1) + 2
|
|
||||||
image_tokens_cumsum.append(
|
|
||||||
image_tokens_cumsum[-1] + num_image_tokens
|
|
||||||
)
|
|
||||||
# text[i] = text[i].replace(self.image_token, self.im_start_token + self.placeholder_token * num_image_tokens + self.im_end_token, 1) # noqa: E501
|
|
||||||
text[i] = text[i].replace(
|
|
||||||
self.image_token, self.placeholder_token * num_image_tokens, 1
|
|
||||||
)
|
|
||||||
index += 1
|
|
||||||
text[i] = text[i].replace(self.placeholder_token, self.image_token)
|
|
||||||
# text[i] = self.tokenizer.bos_token + text[i]
|
|
||||||
|
|
||||||
text_inputs = self.tokenizer(text, add_special_tokens=False, **kwargs)
|
|
||||||
self._check_special_mm_tokens(text, text_inputs, modalities=["image"])
|
|
||||||
|
|
||||||
input_ids = text_inputs["input_ids"]
|
|
||||||
position_ids = torch.arange(len(input_ids[0]))
|
|
||||||
position_ids_w = torch.arange(len(input_ids[0]))
|
|
||||||
position_ids_h = torch.arange(len(input_ids[0]))
|
|
||||||
position_ids_t = torch.arange(len(input_ids[0]))
|
|
||||||
|
|
||||||
if images is not None:
|
|
||||||
image_token_pos_indices = torch.where(input_ids[0] == self.image_token_id)[
|
|
||||||
0
|
|
||||||
]
|
|
||||||
for i in range(len(image_grid_thw)):
|
|
||||||
grid_h, grid_w = image_grid_thw[i][-2:]
|
|
||||||
patch_h = grid_h // self.image_processor.merge_size
|
|
||||||
patch_w = grid_w // self.image_processor.merge_size
|
|
||||||
start_pos = image_token_pos_indices[image_tokens_cumsum[i]].item() + 1
|
|
||||||
replace_num = (patch_w + 1) * patch_h
|
|
||||||
position_ids_w[start_pos : start_pos + replace_num] = torch.tensor(
|
|
||||||
list(range(patch_w + 1)) * patch_h, dtype=torch.int64
|
|
||||||
)
|
|
||||||
patch_h_list = []
|
|
||||||
for h in range(patch_h):
|
|
||||||
patch_h_list += [h] * (patch_w + 1)
|
|
||||||
position_ids_h[start_pos : start_pos + replace_num] = torch.tensor(
|
|
||||||
patch_h_list, dtype=torch.int64
|
|
||||||
)
|
|
||||||
position_ids_t[start_pos : start_pos + replace_num] = 0
|
|
||||||
|
|
||||||
position_ids = torch.stack(
|
|
||||||
[position_ids, position_ids_w, position_ids_h, position_ids_t]
|
|
||||||
).unsqueeze(0)
|
|
||||||
text_inputs["position_ids"] = position_ids
|
|
||||||
|
|
||||||
attention_mask = input_ids.ne(self.pad_id)
|
|
||||||
text_inputs["attention_mask"] = attention_mask
|
|
||||||
text_inputs["imgs_pos"] = [self.get_imgs_pos(e) for e in input_ids]
|
|
||||||
# image_inputs["imgs"] = [[image_inputs["pixel_values"]]]
|
|
||||||
|
|
||||||
return_tensors = kwargs.pop("return_tensors", None)
|
|
||||||
return BatchFeature(
|
|
||||||
data={**text_inputs, **image_inputs},
|
|
||||||
tensor_type=return_tensors,
|
|
||||||
)
|
|
||||||
|
|
||||||
def batch_decode(self, *args, **kwargs):
|
|
||||||
return self.tokenizer.batch_decode(*args, **kwargs)
|
|
||||||
|
|
||||||
def decode(self, *args, **kwargs):
|
|
||||||
return self.tokenizer.decode(*args, **kwargs)
|
|
||||||
|
|
||||||
def post_process_image_text_to_text(
|
|
||||||
self,
|
|
||||||
generated_outputs,
|
|
||||||
skip_special_tokens=True,
|
|
||||||
clean_up_tokenization_spaces=False,
|
|
||||||
**kwargs,
|
|
||||||
):
|
|
||||||
assert 0
|
|
||||||
|
|
||||||
def apply_chat_template(self, *args, **kwargs):
|
|
||||||
kwargs["return_dict"] = False
|
|
||||||
return self.tokenizer.apply_chat_template(*args, **kwargs)
|
|
||||||
|
|
||||||
def get_imgs_pos(self, doc_ids):
|
|
||||||
doc_ids = np.array(doc_ids, dtype=np.int64)
|
|
||||||
img_begin_index = np.where(doc_ids == self.im_start_token_id)[0]
|
|
||||||
img_end_index = np.where(doc_ids == self.im_end_token_id)[0]
|
|
||||||
imgs_pos = np.concatenate(
|
|
||||||
(
|
|
||||||
np.reshape(img_begin_index + 1, (-1, 1)),
|
|
||||||
np.reshape(img_end_index, (-1, 1)),
|
|
||||||
),
|
|
||||||
axis=-1,
|
|
||||||
).tolist()
|
|
||||||
return imgs_pos
|
|
||||||
|
|
||||||
@property
|
|
||||||
def model_input_names(self):
|
|
||||||
tokenizer_input_names = self.tokenizer.model_input_names
|
|
||||||
image_processor_input_names = self.image_processor.model_input_names
|
|
||||||
return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))
|
|
||||||
|
|
||||||
|
|
||||||
def split_image_into_patch_blocks(
|
|
||||||
pixel_values: torch.Tensor, # shape: [batch_size, 3, H, W]
|
|
||||||
patch_size: int = 16, # e.g. 16
|
|
||||||
adaptor_patch_div: int = 4, # e.g. 4 --> each patch_size is cut into 4x4 small regions, i.e. patch_size // 4 # noqa: E501
|
|
||||||
) -> torch.Tensor:
|
|
||||||
"""
|
|
||||||
Split the input image tensor (supporting batch) into large patches of size `patch_size`,
|
|
||||||
and then further divide each large patch into smaller regions of size
|
|
||||||
(patch_size // adaptor_patch_div) x (patch_size // adaptor_patch_div).
|
|
||||||
Each small region is extracted as a tensor of shape [3, patch_size, patch_size].
|
|
||||||
The final output contains all such small region tensors.
|
|
||||||
|
|
||||||
Args:
|
|
||||||
pixel_values: Input image tensor of shape [batch_size, 3, H, W].
|
|
||||||
patch_size: Size of the large patch, e.g., 16.
|
|
||||||
adaptor_patch_div: Each large patch is divided into
|
|
||||||
(patch_size // adaptor_patch_div) x (patch_size // adaptor_patch_div)
|
|
||||||
smaller regions.
|
|
||||||
|
|
||||||
Returns:
|
|
||||||
patches: A tensor of shape [N, 3, patch_size, patch_size],
|
|
||||||
where N = batch_size * (H // patch_size) * (W // patch_size) * (patch_size // adaptor_patch_div)^2.
|
|
||||||
Each element in the batch corresponds to one small image region.
|
|
||||||
""" # noqa: E501
|
|
||||||
batch_size, channels, height, width = pixel_values.shape
|
|
||||||
assert channels == 3, "Pixel values must have 3 channels in dim=1"
|
|
||||||
assert height % patch_size == 0 and width % patch_size == 0, (
|
|
||||||
"H and W must be divisible by patch_size"
|
|
||||||
)
|
|
||||||
|
|
||||||
patch_height_num = height // patch_size
|
|
||||||
patch_width_num = width // patch_size
|
|
||||||
|
|
||||||
# Reshape to [B, 3, ph, ps, pw, ps]
|
|
||||||
img = pixel_values.reshape(
|
|
||||||
batch_size, 3, patch_height_num, patch_size, patch_width_num, patch_size
|
|
||||||
)
|
|
||||||
|
|
||||||
# Further split each psxps patch into (ps//aps)x(ps//aps) small regions
|
|
||||||
img = img.reshape(
|
|
||||||
batch_size,
|
|
||||||
3,
|
|
||||||
patch_height_num,
|
|
||||||
patch_size // adaptor_patch_div, # ps // aps
|
|
||||||
adaptor_patch_div,
|
|
||||||
patch_width_num,
|
|
||||||
patch_size // adaptor_patch_div, # ps // aps
|
|
||||||
adaptor_patch_div,
|
|
||||||
)
|
|
||||||
|
|
||||||
# Permute to group the small regions: [B, ph, pw, ps//aps, ps//aps, 3, aps, aps]
|
|
||||||
img = img.permute(0, 2, 5, 3, 6, 1, 4, 7)
|
|
||||||
|
|
||||||
# Reshape into [B * ph * pw * (ps//aps)^2, 3, patch_size, patch_size]
|
|
||||||
patches = img.reshape(-1, 3, patch_size, patch_size)
|
|
||||||
|
|
||||||
return patches
|
|
||||||
@@ -1,477 +0,0 @@
|
|||||||
# SPDX-License-Identifier: Apache-2.0
|
|
||||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
|
||||||
|
|
||||||
# adapted from https://github.com/ManaEstras/transformers/blob/v4.57.1.hyvl/src/transformers/models/hunyuan_vl/image_processing_hunyuan_vl.py
|
|
||||||
"""Image processor class for HunYuanVL."""
|
|
||||||
|
|
||||||
# isort conflicts with ruff for transformers imports
|
|
||||||
# isort: skip_file
|
|
||||||
import math
|
|
||||||
|
|
||||||
import numpy as np
|
|
||||||
import torchvision.transforms as transforms
|
|
||||||
from transformers import AutoImageProcessor
|
|
||||||
from transformers.image_processing_utils import BaseImageProcessor, BatchFeature
|
|
||||||
from transformers.image_transforms import (
|
|
||||||
convert_to_rgb,
|
|
||||||
)
|
|
||||||
from transformers.image_utils import (
|
|
||||||
OPENAI_CLIP_MEAN,
|
|
||||||
OPENAI_CLIP_STD,
|
|
||||||
ChannelDimension,
|
|
||||||
ImageInput,
|
|
||||||
PILImageResampling,
|
|
||||||
make_flat_list_of_images,
|
|
||||||
make_list_of_images,
|
|
||||||
valid_images,
|
|
||||||
validate_preprocess_arguments,
|
|
||||||
)
|
|
||||||
from transformers.utils import TensorType, logging
|
|
||||||
from transformers.video_utils import VideoInput, make_batched_videos
|
|
||||||
|
|
||||||
logger = logging.get_logger(__name__)
|
|
||||||
|
|
||||||
|
|
||||||
def smart_resize(
|
|
||||||
height: int,
|
|
||||||
width: int,
|
|
||||||
factor: int = 16,
|
|
||||||
min_pixels: int = 512 * 512,
|
|
||||||
max_pixels: int = 2048 * 2048,
|
|
||||||
):
|
|
||||||
"""Rescales the image so that the following conditions are met:
|
|
||||||
|
|
||||||
1. Both dimensions (height and width) are divisible by 'factor'.
|
|
||||||
|
|
||||||
2. The total number of pixels is within the range ['min_pixels', 'max_pixels'].
|
|
||||||
|
|
||||||
3. The aspect ratio of the image is maintained as closely as possible.
|
|
||||||
|
|
||||||
"""
|
|
||||||
if max(height, width) / min(height, width) > 200:
|
|
||||||
raise ValueError(
|
|
||||||
"absolute aspect ratio must be smaller than 200, got "
|
|
||||||
f"{max(height, width) / min(height, width)}"
|
|
||||||
)
|
|
||||||
h_bar = round(height / factor) * factor
|
|
||||||
w_bar = round(width / factor) * factor
|
|
||||||
if h_bar * w_bar > max_pixels:
|
|
||||||
beta = math.sqrt((height * width) / max_pixels)
|
|
||||||
h_bar = max(factor, math.floor(height / beta / factor) * factor)
|
|
||||||
w_bar = max(factor, math.floor(width / beta / factor) * factor)
|
|
||||||
elif h_bar * w_bar < min_pixels:
|
|
||||||
beta = math.sqrt(min_pixels / (height * width))
|
|
||||||
h_bar = math.ceil(height * beta / factor) * factor
|
|
||||||
w_bar = math.ceil(width * beta / factor) * factor
|
|
||||||
return h_bar, w_bar
|
|
||||||
|
|
||||||
|
|
||||||
class HunYuanVLImageProcessor(BaseImageProcessor):
|
|
||||||
model_input_names = [
|
|
||||||
"pixel_values",
|
|
||||||
"image_grid_thw",
|
|
||||||
"pixel_values_videos",
|
|
||||||
"video_grid_thw",
|
|
||||||
]
|
|
||||||
|
|
||||||
def __init__(
|
|
||||||
self,
|
|
||||||
do_resize: bool = True,
|
|
||||||
size: dict[str, int] | None = None,
|
|
||||||
resample: PILImageResampling = PILImageResampling.BICUBIC,
|
|
||||||
do_rescale: bool = True,
|
|
||||||
rescale_factor: int | float = 1 / 255,
|
|
||||||
do_normalize: bool = True,
|
|
||||||
image_mean: float | list[float] | None = None,
|
|
||||||
image_std: float | list[float] | None = None,
|
|
||||||
do_convert_rgb: bool = True,
|
|
||||||
min_pixels: int | None = None,
|
|
||||||
max_pixels: int | None = None,
|
|
||||||
patch_size: int = 16,
|
|
||||||
temporal_patch_size: int = 2,
|
|
||||||
merge_size: int = 2,
|
|
||||||
**kwargs,
|
|
||||||
) -> None:
|
|
||||||
super().__init__(**kwargs)
|
|
||||||
if size is not None and (
|
|
||||||
"shortest_edge" not in size or "longest_edge" not in size
|
|
||||||
):
|
|
||||||
raise ValueError(
|
|
||||||
"size must contain 'shortest_edge' and 'longest_edge' keys."
|
|
||||||
)
|
|
||||||
else:
|
|
||||||
size = {"shortest_edge": 512 * 512, "longest_edge": 2048 * 2048}
|
|
||||||
# backward compatibility: override size with min_pixels and max_pixels
|
|
||||||
# if they are provided.
|
|
||||||
if min_pixels is not None:
|
|
||||||
size["shortest_edge"] = min_pixels
|
|
||||||
if max_pixels is not None:
|
|
||||||
size["longest_edge"] = max_pixels
|
|
||||||
self.min_pixels = size["shortest_edge"]
|
|
||||||
self.max_pixels = size["longest_edge"]
|
|
||||||
self.size = size
|
|
||||||
|
|
||||||
self.do_resize = do_resize
|
|
||||||
self.resample = resample
|
|
||||||
self.do_rescale = do_rescale
|
|
||||||
self.rescale_factor = rescale_factor
|
|
||||||
self.do_normalize = do_normalize
|
|
||||||
self.image_mean = image_mean if image_mean is not None else OPENAI_CLIP_MEAN
|
|
||||||
self.image_std = image_std if image_std is not None else OPENAI_CLIP_STD
|
|
||||||
|
|
||||||
self.patch_size = patch_size
|
|
||||||
self.temporal_patch_size = temporal_patch_size
|
|
||||||
self.merge_size = merge_size
|
|
||||||
self.do_convert_rgb = do_convert_rgb
|
|
||||||
|
|
||||||
# hard-code
|
|
||||||
|
|
||||||
def _preprocess(
|
|
||||||
self,
|
|
||||||
images: ImageInput | VideoInput,
|
|
||||||
do_resize: bool | None = None,
|
|
||||||
size: dict[str, int] | None = None,
|
|
||||||
resample: PILImageResampling = None,
|
|
||||||
do_rescale: bool | None = None,
|
|
||||||
rescale_factor: float | None = None,
|
|
||||||
do_normalize: bool | None = None,
|
|
||||||
image_mean: float | list[float] | None = None,
|
|
||||||
image_std: float | list[float] | None = None,
|
|
||||||
patch_size: int = 16,
|
|
||||||
temporal_patch_size: int = 2,
|
|
||||||
merge_size: int = 2,
|
|
||||||
do_convert_rgb: bool | None = None,
|
|
||||||
data_format: ChannelDimension | None = ChannelDimension.FIRST,
|
|
||||||
input_data_format: str | ChannelDimension | None = None,
|
|
||||||
):
|
|
||||||
"""
|
|
||||||
Preprocess an image or batch of images. Copy of the `preprocess` method from `CLIPImageProcessor`.
|
|
||||||
|
|
||||||
Args:
|
|
||||||
images (`ImageInput`):
|
|
||||||
Image or batch of images to preprocess. Expects pixel values ranging from 0 to 255. If pixel values range from 0 to 1, set `do_rescale=False`.
|
|
||||||
do_resize (`bool`, *optional*, defaults to `self.do_resize`):
|
|
||||||
Whether to resize the image.
|
|
||||||
size (`dict[str, int]`, *optional*, defaults to `self.size`):
|
|
||||||
Size of the image after resizing. `shortest_edge` and `longest_edge` keys must be present.
|
|
||||||
resample (`PILImageResampling`, *optional*, defaults to `self.resample`):
|
|
||||||
Resampling filter to use if resizing the image. This can be one of the `PILImageResampling` enums.
|
|
||||||
do_rescale (`bool`, *optional*, defaults to `self.do_rescale`):
|
|
||||||
Whether to rescale the image.
|
|
||||||
rescale_factor (`float`, *optional*, defaults to `self.rescale_factor`):
|
|
||||||
Scale factor to use if rescaling the image.
|
|
||||||
do_normalize (`bool`, *optional*, defaults to `self.do_normalize`):
|
|
||||||
Whether to normalize the image.
|
|
||||||
image_mean (`float` or `list[float]`, *optional*, defaults to `self.image_mean`):
|
|
||||||
Mean to use if normalizing the image. Can be a float or a list of floats corresponding to the number of channels in the image.
|
|
||||||
image_std (`float` or `list[float]`, *optional*, defaults to `self.image_std`):
|
|
||||||
Standard deviation to use if normalizing the image. Can be a float or a list of floats corresponding to the number of channels in the image.
|
|
||||||
patch_size (`int`, *optional*, defaults to `self.patch_size`):
|
|
||||||
The spatial patch size of the vision encoder.
|
|
||||||
temporal_patch_size (`int`, *optional*, defaults to `self.temporal_patch_size`):
|
|
||||||
The temporal patch size of the vision encoder.
|
|
||||||
merge_size (`int`, *optional*, defaults to `self.merge_size`):
|
|
||||||
The merge size of the vision encoder to llm encoder.
|
|
||||||
do_convert_rgb (`bool`, *optional*, defaults to `self.do_convert_rgb`):
|
|
||||||
Whether to convert the image to RGB.
|
|
||||||
data_format (`ChannelDimension`, *optional*, defaults to `ChannelDimension.FIRST`):
|
|
||||||
The channel dimension format for the output image. Can be one of:
|
|
||||||
- `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
|
|
||||||
- `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
|
|
||||||
- Unset: Use the channel dimension format of the input image.
|
|
||||||
input_data_format (`ChannelDimension` or `str`, *optional*):
|
|
||||||
The channel dimension format for the input image. Can be one of:
|
|
||||||
- `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
|
|
||||||
- `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
|
|
||||||
- `"none"` or `ChannelDimension.NONE`: image in (height, width) format. - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
|
|
||||||
""" # noqa: E501
|
|
||||||
images = make_list_of_images(images)
|
|
||||||
|
|
||||||
if do_convert_rgb:
|
|
||||||
images = [convert_to_rgb(image) for image in images]
|
|
||||||
|
|
||||||
width, height = images[0].width, images[0].height
|
|
||||||
resized_width, resized_height = width, height
|
|
||||||
processed_images = []
|
|
||||||
for image in images:
|
|
||||||
if do_resize:
|
|
||||||
resized_height, resized_width = smart_resize(
|
|
||||||
height=height,
|
|
||||||
width=width,
|
|
||||||
factor=patch_size * merge_size,
|
|
||||||
min_pixels=self.min_pixels,
|
|
||||||
max_pixels=self.max_pixels,
|
|
||||||
)
|
|
||||||
image = image.resize((resized_width, resized_height))
|
|
||||||
|
|
||||||
if do_normalize:
|
|
||||||
image = transforms.Compose(
|
|
||||||
[
|
|
||||||
transforms.ToTensor(),
|
|
||||||
transforms.Normalize(self.image_mean, self.image_std),
|
|
||||||
]
|
|
||||||
)(image)
|
|
||||||
processed_images.append(image)
|
|
||||||
|
|
||||||
patches = np.array(processed_images)
|
|
||||||
channel = patches.shape[1]
|
|
||||||
grid_t = patches.shape[0] // temporal_patch_size
|
|
||||||
grid_h, grid_w = resized_height // patch_size, resized_width // patch_size
|
|
||||||
patches = patches.reshape(
|
|
||||||
1,
|
|
||||||
channel,
|
|
||||||
grid_h // merge_size,
|
|
||||||
merge_size,
|
|
||||||
patch_size,
|
|
||||||
grid_w // merge_size,
|
|
||||||
merge_size,
|
|
||||||
patch_size,
|
|
||||||
)
|
|
||||||
patches = patches.transpose(0, 2, 3, 5, 6, 1, 4, 7)
|
|
||||||
flatten_patches = patches.reshape(
|
|
||||||
1 * grid_h * grid_w, channel * patch_size * patch_size
|
|
||||||
)
|
|
||||||
|
|
||||||
return flatten_patches, (grid_t, grid_h, grid_w)
|
|
||||||
|
|
||||||
def preprocess(
|
|
||||||
self,
|
|
||||||
images: ImageInput,
|
|
||||||
videos: VideoInput = None,
|
|
||||||
do_resize: bool | None = None,
|
|
||||||
size: dict[str, int] | None = None,
|
|
||||||
min_pixels: int | None = None,
|
|
||||||
max_pixels: int | None = None,
|
|
||||||
resample: PILImageResampling = None,
|
|
||||||
do_rescale: bool | None = None,
|
|
||||||
rescale_factor: float | None = None,
|
|
||||||
do_normalize: bool | None = None,
|
|
||||||
image_mean: float | list[float] | None = None,
|
|
||||||
image_std: float | list[float] | None = None,
|
|
||||||
patch_size: int | None = None,
|
|
||||||
temporal_patch_size: int | None = None,
|
|
||||||
merge_size: int | None = None,
|
|
||||||
do_convert_rgb: bool | None = None,
|
|
||||||
return_tensors: str | TensorType | None = None,
|
|
||||||
data_format: ChannelDimension | None = ChannelDimension.FIRST,
|
|
||||||
input_data_format: str | ChannelDimension | None = None,
|
|
||||||
):
|
|
||||||
"""
|
|
||||||
Args:
|
|
||||||
images (`ImageInput`):
|
|
||||||
Image to preprocess. Expects a single or batch of images with pixel values ranging from 0 to 255. If
|
|
||||||
passing in images with pixel values between 0 and 1, set `do_rescale=False`.
|
|
||||||
videos (`VideoInput`):
|
|
||||||
Video to preprocess. Expects a single or batch of videos with pixel values ranging from 0 to 255. If
|
|
||||||
passing in videos with pixel values between 0 and 1, set `do_rescale=False`.
|
|
||||||
do_resize (`bool`, *optional*, defaults to `self.do_resize`):
|
|
||||||
Whether to resize the image.
|
|
||||||
size (`dict[str, int]`, *optional*, defaults to `self.size`):
|
|
||||||
Size of the image after resizing. Shortest edge of the image is resized to size["shortest_edge"], with
|
|
||||||
the longest edge resized to keep the input aspect ratio.
|
|
||||||
resample (`int`, *optional*, defaults to `self.resample`):
|
|
||||||
Resampling filter to use if resizing the image. This can be one of the enum `PILImageResampling`. Only
|
|
||||||
has an effect if `do_resize` is set to `True`.
|
|
||||||
do_rescale (`bool`, *optional*, defaults to `self.do_rescale`):
|
|
||||||
Whether to rescale the image.
|
|
||||||
rescale_factor (`float`, *optional*, defaults to `self.rescale_factor`):
|
|
||||||
Rescale factor to rescale the image by if `do_rescale` is set to `True`.
|
|
||||||
do_normalize (`bool`, *optional*, defaults to `self.do_normalize`):
|
|
||||||
Whether to normalize the image.
|
|
||||||
image_mean (`float` or `list[float]`, *optional*, defaults to `self.image_mean`):
|
|
||||||
Image mean to use for normalization. Only has an effect if `do_normalize` is set to `True`.
|
|
||||||
image_std (`float` or `list[float]`, *optional*, defaults to `self.image_std`):
|
|
||||||
Image standard deviation to use for normalization. Only has an effect if `do_normalize` is set to
|
|
||||||
`True`.
|
|
||||||
min_pixels (`int`, *optional*, defaults to `self.min_pixels`):
|
|
||||||
The min pixels of the image to resize the image.
|
|
||||||
max_pixels (`int`, *optional*, defaults to `self.max_pixels`):
|
|
||||||
The max pixels of the image to resize the image.
|
|
||||||
patch_size (`int`, *optional*, defaults to `self.patch_size`):
|
|
||||||
The spatial patch size of the vision encoder.
|
|
||||||
temporal_patch_size (`int`, *optional*, defaults to `self.temporal_patch_size`):
|
|
||||||
The temporal patch size of the vision encoder.
|
|
||||||
merge_size (`int`, *optional*, defaults to `self.merge_size`):
|
|
||||||
The merge size of the vision encoder to llm encoder.
|
|
||||||
do_convert_rgb (`bool`, *optional*, defaults to `self.do_convert_rgb`):
|
|
||||||
Whether to convert the image to RGB.
|
|
||||||
return_tensors (`str` or `TensorType`, *optional*):
|
|
||||||
The type of tensors to return. Can be one of:
|
|
||||||
- Unset: Return a list of `np.ndarray`.
|
|
||||||
- `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`.
|
|
||||||
- `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
|
|
||||||
- `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
|
|
||||||
- `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`.
|
|
||||||
data_format (`ChannelDimension` or `str`, *optional*, defaults to `ChannelDimension.FIRST`):
|
|
||||||
The channel dimension format for the output image. Can be one of:
|
|
||||||
- `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
|
|
||||||
- `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
|
|
||||||
- Unset: Use the channel dimension format of the input image.
|
|
||||||
input_data_format (`ChannelDimension` or `str`, *optional*):
|
|
||||||
The channel dimension format for the input image. If unset, the channel dimension format is inferred
|
|
||||||
from the input image. Can be one of:
|
|
||||||
- `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
|
|
||||||
- `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
|
|
||||||
- `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
|
|
||||||
|
|
||||||
""" # noqa: E501
|
|
||||||
min_pixels = min_pixels if min_pixels is not None else self.min_pixels
|
|
||||||
max_pixels = max_pixels if max_pixels is not None else self.max_pixels
|
|
||||||
|
|
||||||
if size is not None:
|
|
||||||
if "shortest_edge" not in size or "longest_edge" not in size:
|
|
||||||
raise ValueError(
|
|
||||||
"size must contain 'shortest_edge' and 'longest_edge' keys."
|
|
||||||
)
|
|
||||||
min_pixels = size["shortest_edge"]
|
|
||||||
elif min_pixels is not None and max_pixels is not None:
|
|
||||||
# backward compatibility: override size with min_pixels and max_pixels
|
|
||||||
# if they are provided.
|
|
||||||
size = {"shortest_edge": min_pixels, "longest_edge": max_pixels}
|
|
||||||
else:
|
|
||||||
size = {**self.size}
|
|
||||||
|
|
||||||
do_resize = do_resize if do_resize is not None else self.do_resize
|
|
||||||
|
|
||||||
resample = resample if resample is not None else self.resample
|
|
||||||
do_rescale = do_rescale if do_rescale is not None else self.do_rescale
|
|
||||||
rescale_factor = (
|
|
||||||
rescale_factor if rescale_factor is not None else self.rescale_factor
|
|
||||||
)
|
|
||||||
do_normalize = do_normalize if do_normalize is not None else self.do_normalize
|
|
||||||
image_mean = image_mean if image_mean is not None else self.image_mean
|
|
||||||
image_std = image_std if image_std is not None else self.image_std
|
|
||||||
patch_size = patch_size if patch_size is not None else self.patch_size
|
|
||||||
temporal_patch_size = (
|
|
||||||
temporal_patch_size
|
|
||||||
if temporal_patch_size is not None
|
|
||||||
else self.temporal_patch_size
|
|
||||||
)
|
|
||||||
merge_size = merge_size if merge_size is not None else self.merge_size
|
|
||||||
do_convert_rgb = (
|
|
||||||
do_convert_rgb if do_convert_rgb is not None else self.do_convert_rgb
|
|
||||||
)
|
|
||||||
|
|
||||||
if images is not None:
|
|
||||||
images = make_flat_list_of_images(images)
|
|
||||||
|
|
||||||
if images is not None and not valid_images(images):
|
|
||||||
raise ValueError(
|
|
||||||
"Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, "
|
|
||||||
"torch.Tensor, tf.Tensor or jax.ndarray."
|
|
||||||
)
|
|
||||||
|
|
||||||
validate_preprocess_arguments(
|
|
||||||
rescale_factor=rescale_factor,
|
|
||||||
do_normalize=do_normalize,
|
|
||||||
image_mean=image_mean,
|
|
||||||
image_std=image_std,
|
|
||||||
do_resize=do_resize,
|
|
||||||
size=size,
|
|
||||||
resample=resample,
|
|
||||||
)
|
|
||||||
|
|
||||||
data = {}
|
|
||||||
if images is not None:
|
|
||||||
pixel_values, vision_grid_thws = [], []
|
|
||||||
for image in images:
|
|
||||||
patches, image_grid_thw = self._preprocess(
|
|
||||||
image,
|
|
||||||
do_resize=do_resize,
|
|
||||||
size=size,
|
|
||||||
resample=resample,
|
|
||||||
do_rescale=do_rescale,
|
|
||||||
rescale_factor=rescale_factor,
|
|
||||||
do_normalize=do_normalize,
|
|
||||||
image_mean=image_mean,
|
|
||||||
image_std=image_std,
|
|
||||||
patch_size=patch_size,
|
|
||||||
temporal_patch_size=temporal_patch_size,
|
|
||||||
merge_size=merge_size,
|
|
||||||
data_format=data_format,
|
|
||||||
do_convert_rgb=do_convert_rgb,
|
|
||||||
input_data_format=input_data_format,
|
|
||||||
)
|
|
||||||
pixel_values.extend(patches)
|
|
||||||
vision_grid_thws.append(image_grid_thw)
|
|
||||||
pixel_values = np.array(pixel_values)
|
|
||||||
vision_grid_thws = np.array(vision_grid_thws)
|
|
||||||
data.update(
|
|
||||||
{"pixel_values": pixel_values, "image_grid_thw": vision_grid_thws}
|
|
||||||
)
|
|
||||||
|
|
||||||
# kept for BC only and should be removed after v5.0
|
|
||||||
if videos is not None:
|
|
||||||
logger.warning(
|
|
||||||
"`HunYuanVLV1ImageProcessor` works only with image inputs "
|
|
||||||
"and doesn't process videos anymore. "
|
|
||||||
"This is a deprecated behavior and will be removed in v5.0. "
|
|
||||||
"Your videos should be forwarded to `HunYuanVLV1VideoProcessor`. "
|
|
||||||
)
|
|
||||||
videos = make_batched_videos(videos)
|
|
||||||
pixel_values_videos, vision_grid_thws_videos = [], []
|
|
||||||
for images in videos:
|
|
||||||
patches, video_grid_thw = self._preprocess(
|
|
||||||
images,
|
|
||||||
do_resize=do_resize,
|
|
||||||
size=size,
|
|
||||||
resample=resample,
|
|
||||||
do_rescale=do_rescale,
|
|
||||||
rescale_factor=rescale_factor,
|
|
||||||
do_normalize=do_normalize,
|
|
||||||
image_mean=image_mean,
|
|
||||||
image_std=image_std,
|
|
||||||
patch_size=patch_size,
|
|
||||||
temporal_patch_size=temporal_patch_size,
|
|
||||||
merge_size=merge_size,
|
|
||||||
data_format=data_format,
|
|
||||||
do_convert_rgb=do_convert_rgb,
|
|
||||||
input_data_format=input_data_format,
|
|
||||||
)
|
|
||||||
pixel_values_videos.extend(patches)
|
|
||||||
vision_grid_thws_videos.append(video_grid_thw)
|
|
||||||
data.update(
|
|
||||||
{
|
|
||||||
"pixel_values_videos": np.array(pixel_values_videos),
|
|
||||||
"video_grid_thw": np.array(vision_grid_thws_videos),
|
|
||||||
}
|
|
||||||
)
|
|
||||||
|
|
||||||
return BatchFeature(data=data, tensor_type=return_tensors)
|
|
||||||
|
|
||||||
def get_number_of_image_patches(self, height: int, width: int, images_kwargs=None):
|
|
||||||
"""
|
|
||||||
A utility that returns number of image patches for a given image size.
|
|
||||||
|
|
||||||
Args:
|
|
||||||
height (`int`):
|
|
||||||
Height of the input image.
|
|
||||||
width (`int`):
|
|
||||||
Width of the input image.
|
|
||||||
images_kwargs (`dict`, *optional*):
|
|
||||||
Any kwargs to override defaults of the image processor.
|
|
||||||
Returns:
|
|
||||||
`int`: Number of image patches per image.
|
|
||||||
"""
|
|
||||||
min_pixels = (
|
|
||||||
images_kwargs["min_pixels"]
|
|
||||||
if "min_pixels" in images_kwargs
|
|
||||||
else self.size["shortest_edge"]
|
|
||||||
)
|
|
||||||
max_pixels = (
|
|
||||||
images_kwargs["max_pixels"]
|
|
||||||
if "max_pixels" in images_kwargs
|
|
||||||
else self.size["longest_edge"]
|
|
||||||
)
|
|
||||||
patch_size = images_kwargs.get("patch_size", self.patch_size)
|
|
||||||
merge_size = images_kwargs.get("merge_size", self.merge_size)
|
|
||||||
|
|
||||||
factor = patch_size * merge_size
|
|
||||||
resized_height, resized_width = smart_resize(
|
|
||||||
height, width, factor, min_pixels=min_pixels, max_pixels=max_pixels
|
|
||||||
)
|
|
||||||
grid_h, grid_w = resized_height // patch_size, resized_width // patch_size
|
|
||||||
return grid_h * (grid_w + 1) + 2
|
|
||||||
|
|
||||||
|
|
||||||
AutoImageProcessor.register("HunYuanVLImageProcessor", HunYuanVLImageProcessor)
|
|
||||||
Reference in New Issue
Block a user