Upload 8 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

chat_template.jinja

@@ -0,0 +1,43 @@
+{%- if not add_generation_prompt is defined -%}
+    {%- set add_generation_prompt = true -%}
+{%- endif -%}
+{%- if not cls_token is defined -%}
+    {%- set cls_token = "<|begin_of_sentence|>" -%}
+{%- endif -%}
+{%- if not eos_token is defined -%}
+    {%- set eos_token = "</s>" -%}
+{%- endif -%}
+{{- cls_token -}}
+{%- for message in messages -%}
+    {%- if message["role"] == "user" -%}
+        {{- "User: " -}}
+        {%- for content in message["content"] -%}
+            {%- if content["type"] == "image" -%}
+                {{ "<|IMAGE_START|><|IMAGE_PLACEHOLDER|><|IMAGE_END|>" }}
+            {%- endif -%}
+        {%- endfor -%}
+        {%- for content in message["content"] -%}
+            {%- if content["type"] == "text" -%}
+                {{ content["text"] }}
+            {%- endif -%}
+        {%- endfor -%}
+        {{ "\n" -}}
+    {%- elif message["role"] == "assistant" -%}
+        {{- "Assistant: " -}}
+        {%- for content in message["content"] -%}
+            {%- if content["type"] == "text" -%}
+                {{ content["text"] }}
+            {%- endif -%}
+        {%- endfor -%}
+        {{ eos_token -}}
+    {%- elif message["role"] == "system" -%}
+        {%- for content in message["content"] -%}
+            {%- if content["type"] == "text" -%}
+                {{ content["text"] + "\n" }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- endif -%}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{- "Assistant: " -}}
+{%- endif -%}

configuration_paddleocr_vl.py

@@ -0,0 +1,191 @@
+# Copyright (c) 2025 PaddlePaddle Authors. 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 transformers.configuration_utils import PretrainedConfig
+from transformers.modeling_rope_utils import rope_config_validation
+
+class PaddleOCRVisionConfig(PretrainedConfig):
+    model_type = "paddleocr_vl"
+    base_config_key = "vision_config"
+
+    def __init__(
+        self,
+        hidden_size=768,
+        intermediate_size=3072,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+        num_channels=3,
+        image_size=224,
+        patch_size=14,
+        hidden_act="gelu_pytorch_tanh",
+        layer_norm_eps=1e-6,
+        attention_dropout=0.0,
+        spatial_merge_size=2,
+        temporal_patch_size=2,
+        tokens_per_second=2,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_channels = num_channels
+        self.patch_size = patch_size
+        self.image_size = image_size
+        self.attention_dropout = attention_dropout
+        self.layer_norm_eps = layer_norm_eps
+        self.hidden_act = hidden_act
+        self.spatial_merge_size = spatial_merge_size
+        self.temporal_patch_size = temporal_patch_size
+        self.tokens_per_second = tokens_per_second
+
+
+
+class PaddleOCRVLConfig(PretrainedConfig):
+    """
+    Configuration class.
+
+    This class stores the configuration of an Ernie model, defining the model architecture.
+    It inherits from PretrainedConfig and can be used to control model outputs.
+    """
+
+    model_type = "paddleocr_vl"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    sub_configs = {"vision_config": PaddleOCRVisionConfig}
+
+    # Default tensor parallel plan for base model `Qwen3`
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_proj": "colwise",
+        "layers.*.self_attn.k_proj": "colwise",
+        "layers.*.self_attn.v_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.mlp.gate_proj": "colwise",
+        "layers.*.mlp.up_proj": "colwise",
+        "layers.*.mlp.down_proj": "rowwise",
+    }
+    base_model_pp_plan = {
+        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
+        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
+        "norm": (["hidden_states"], ["hidden_states"]),
+    }
+
+    def __init__(
+        self,
+        vocab_size=32000,
+        hidden_size=768,
+        intermediate_size=11008,
+        max_position_embeddings=32768,
+        num_hidden_layers=2,
+        num_attention_heads=2,
+        image_token_id=101304,
+        video_token_id=101305,
+        vision_start_token_id=101306,
+        rms_norm_eps=1e-6,
+        use_cache=False,
+        use_flash_attention=False,
+        pad_token_id=0,
+        bos_token_id=1,
+        eos_token_id=2,
+        head_dim=128,
+        hidden_act="silu",
+        use_bias=False,
+        rope_theta=10000,
+        weight_share_add_bias=True,
+        ignored_index=-100,
+        attention_probs_dropout_prob=0.0,
+        hidden_dropout_prob=0.0,
+        compression_ratio: float = 1.0,
+        num_key_value_heads=None,
+        max_sequence_length=None,
+        tie_word_embeddings=False,
+        vision_config=None,
+        rope_scaling=None,
+        **kwargs,
+    ):
+        """
+        Initialize configuration with default or specified parameters.
+
+        Args:
+            vocab_size (int): Size of the vocabulary (number of unique tokens)
+            hidden_size (int): Dimensionality of the encoder layers and the pooler layer
+            intermediate_size (int): Dimensionality of the "intermediate" (feed-forward) layer
+            max_position_embeddings (int): Maximum sequence length the model can handle
+            num_hidden_layers (int): Number of hidden layers in the Transformer encoder
+            num_attention_heads (int): Number of attention heads for each attention layer
+            rms_norm_eps (float): The epsilon used by the RMS normalization layers
+            use_cache (bool): Whether to use caching for faster generation (decoding)
+            use_flash_attention (bool): Whether to use FlashAttention for optimized attention computation
+            pad_token_id (int): Token ID used for padding sequences
+            bos_token_id (int): Token ID used for beginning-of-sequence
+            eos_token_id (int): Token ID used for end-of-sequence
+            use_bias (bool): Whether to use bias terms in linear layers
+            rope_theta (float): The base period of the RoPE embeddings
+            weight_share_add_bias (bool): Whether to share bias weights in certain layers
+            ignored_index (int): Target value that is ignored during loss computation
+            attention_probs_dropout_prob (float): Dropout probability for attention weights
+            hidden_dropout_prob (float): Dropout probability for hidden layers
+            compression_ratio (float): Ratio for KV cache compression (1.0 = no compression)
+            num_key_value_heads (int): Number of key/value heads (for Grouped Query Attention)
+            max_sequence_length (int): Maximum sequence length for positional embeddings
+            **kwargs: Additional keyword arguments passed to parent class
+        """
+
+        # Set default for tied embeddings if not specified.
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            **kwargs,
+        )
+        if isinstance(vision_config, dict):
+            self.vision_config = self.sub_configs["vision_config"](**vision_config)
+        elif vision_config is None:
+            self.vision_config = self.sub_configs["vision_config"]()        
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.max_position_embeddings = max_position_embeddings
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.use_flash_attention = use_flash_attention
+        self.pad_token_id = pad_token_id
+        self.bos_token_id = bos_token_id
+        self.eos_token_id = eos_token_id
+        self.image_token_id = image_token_id
+        self.video_token_id = video_token_id
+        self.vision_start_token_id = vision_start_token_id
+        self.head_dim = head_dim
+        self.hidden_act=hidden_act
+        self.sliding_window = None
+        self.hidden_size = hidden_size
+        self.use_bias = use_bias
+        self.weight_share_add_bias = weight_share_add_bias
+        self.rope_theta = rope_theta
+        self.ignored_index = ignored_index
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.compression_ratio = compression_ratio
+        self.num_key_value_heads = num_key_value_heads
+        self.max_sequence_length = max_sequence_length
+        self.rope_scaling = rope_scaling
+        if self.rope_scaling is not None and "type" in self.rope_scaling:
+            if self.rope_scaling["type"] == "mrope":
+                self.rope_scaling["type"] = "default"
+            self.rope_scaling["rope_type"] = self.rope_scaling["type"]
+        rope_config_validation(self, ignore_keys={"mrope_section"})        
+        super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)

image_processing_paddleocr_vl.py

@@ -0,0 +1,570 @@
+# Copyright (c) 2025 PaddlePaddle Authors. 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.
+
+"""Image processor class for PaddleOCR-VL."""
+
+import math
+from typing import Dict, List, Optional, Union
+
+import numpy as np
+import torch
+from transformers.image_processing_utils import BaseImageProcessor, BatchFeature
+from torchvision.transforms import functional as TF
+from transformers.image_transforms import (
+    convert_to_rgb,
+    resize,
+    to_channel_dimension_format,
+)
+from transformers.image_utils import (
+    OPENAI_CLIP_MEAN,
+    OPENAI_CLIP_STD,
+    ChannelDimension,
+    PILImageResampling,
+    get_image_size,
+    infer_channel_dimension_format,
+    is_scaled_image,
+    is_valid_image,
+    make_list_of_images,
+    to_numpy_array,
+    valid_images,
+    validate_preprocess_arguments,
+)
+from transformers.utils import TensorType, is_vision_available, logging
+
+
+logger = logging.get_logger(__name__)
+
+
+if is_vision_available():
+    from PIL import Image
+
+ImageInput = Union[
+    "PIL.Image.Image",
+    np.ndarray,
+    "torch.Tensor",
+    List["PIL.Image.Image"],
+    List[np.ndarray],
+    List["torch.Tensor"],
+]  # noqa
+
+
+VideoInput = Union[
+    List["PIL.Image.Image"],
+    "np.ndarray",
+    "torch.Tensor",
+    List["np.ndarray"],
+    List["torch.Tensor"],
+    List[List["PIL.Image.Image"]],
+    List[List["np.ndarrray"]],
+    List[List["torch.Tensor"]],
+]  # noqa
+
+
+def make_batched_images(images) -> List[List[ImageInput]]:
+    """
+    Accepts images in list or nested list format, and makes a list of images for preprocessing.
+
+    Args:
+        images (`Union[List[List[ImageInput]], List[ImageInput], ImageInput]`):
+            The input image.
+
+    Returns:
+        list: A list of images.
+    """
+    if (
+        isinstance(images, (list, tuple))
+        and isinstance(images[0], (list, tuple))
+        and is_valid_image(images[0][0])
+    ):
+        return [img for img_list in images for img in img_list]
+
+    elif isinstance(images, (list, tuple)) and is_valid_image(images[0]):
+        return images
+
+    elif is_valid_image(images):
+        return [images]
+
+    raise ValueError(f"Could not make batched images from {images}")
+
+
+def adjust_size(size, patch_size):
+    num_patches = size // patch_size
+    if num_patches % 2 != 0:  # 如果是奇数，减1
+        num_patches -= 1
+    return num_patches * patch_size
+
+
+def make_batched_videos(videos) -> List[VideoInput]:
+    if (
+        isinstance(videos, (list, tuple))
+        and isinstance(videos[0], (list, tuple))
+        and is_valid_image(videos[0][0])
+    ):
+        return videos
+
+    elif isinstance(videos, (list, tuple)) and is_valid_image(videos[0]):
+        if isinstance(videos[0], Image.Image):
+            return [videos]
+        elif len(videos[0].shape) == 4:
+            return [list(video) for video in videos]
+
+    elif is_valid_image(videos) and len(videos.shape) == 4:
+        return [list(videos)]
+
+    raise ValueError(f"Could not make batched video from {videos}")
+
+
+def smart_resize(
+    height: int,
+    width: int,
+    factor: int = 28,
+    min_pixels: int = 28 * 28 * 130,
+    max_pixels: int = 28 * 28 * 1280,
+):
+    """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 height < factor or width < factor:
+    #    raise ValueError(f"height:{height} or width:{width} must be larger than factor:{factor}")
+    # if int(height < factor//4) + int(width < factor//4):
+    #     raise ValueError(f"height:{height} or width:{width} must be larger than factor:{factor//4}")
+
+    if height < factor:
+        print(f"smart_resize: height={height} < factor={factor}, reset height=factor")
+        width = round((width * factor) / height)
+        height = factor
+
+    if width < factor:
+        print(f"smart_resize: width={width} < factor={factor}, reset width=factor")
+        height = round((height * factor) / width)
+        width = factor
+
+    if max(height, width) / min(height, width) > 200:
+        raise ValueError(
+            f"absolute aspect ratio must be smaller than 200, got {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 = math.floor(height / beta / factor) * factor
+        w_bar = 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 PaddleOCRVLImageProcessor(BaseImageProcessor):
+    r"""
+    Constructs a Siglip image processor that dynamically resizes images based on the original images.
+
+    Args:
+        do_resize (`bool`, *optional*, defaults to `True`):
+            Whether to resize the image's (height, width) dimensions.
+        resample (`PILImageResampling`, *optional*, defaults to `Resampling.BICUBIC`):
+            Resampling filter to use when resizing the image.
+        do_rescale (`bool`, *optional*, defaults to `True`):
+            Whether to rescale the image by the specified scale `rescale_factor`.
+        rescale_factor (`int` or `float`, *optional*, defaults to `1/255`):
+            Scale factor to use if rescaling the image.
+        do_normalize (`bool`, *optional*, defaults to `True`):
+            Whether to normalize the image.
+        image_mean (`float` or `List[float]`, *optional*, defaults to `[0.48145466, 0.4578275, 0.40821073]`):
+            Mean to use if normalizing the image. This is a float or list of floats for each channel in the image.
+        image_std (`float` or `List[float]`, *optional*, defaults to `[0.26862954, 0.26130258, 0.27577711]`):
+            Standard deviation to use if normalizing the image. This is a float or list of floats for each channel in the image.
+        do_convert_rgb (`bool`, *optional*, defaults to `True`):
+            Whether to convert the image to RGB.
+        min_pixels (`int`, *optional*, defaults to `28 * 28 * 130`):
+            The min pixels of the image to resize the image.
+        max_pixels (`int`, *optional*, defaults to `28 * 28 * 1670`):
+            The max pixels of the image to resize the image.
+        patch_size (`int`, *optional*, defaults to 14):
+            The spacial patch size of the vision encoder.
+        temporal_patch_size (`int`, *optional*, defaults to 2):
+            The temporal patch size of the vision encoder.
+        merge_size (`int`, *optional*, defaults to 2):
+            The merge size of the vision encoder to llm encoder.
+    """
+
+    model_input_names = [
+        "pixel_values",
+        "image_grid_thw",
+        "pixel_values_videos",
+        "video_grid_thw",
+    ]
+
+    def __init__(
+        self,
+        do_resize: bool = True,
+        resample: PILImageResampling = PILImageResampling.BICUBIC,
+        do_rescale: bool = True,
+        rescale_factor: Union[int, float] = 1 / 255,
+        do_normalize: bool = True,
+        image_mean: Optional[Union[float, List[float]]] = None,
+        image_std: Optional[Union[float, List[float]]] = None,
+        do_convert_rgb: bool = True,
+        min_pixels: int = 28 * 28 * 130,
+        max_pixels: int = 28 * 28 * 1280,
+        patch_size: int = 14,
+        temporal_patch_size: int = 1,
+        merge_size: int = 2,
+        **kwargs,
+    ) -> None:
+        super().__init__(**kwargs)
+        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.min_pixels = min_pixels
+        self.max_pixels = max_pixels
+        self.patch_size = patch_size
+        self.temporal_patch_size = temporal_patch_size
+        self.merge_size = merge_size
+        self.size = {"min_pixels": min_pixels, "max_pixels": max_pixels}  # not used
+        self.do_convert_rgb = do_convert_rgb
+
+    def mvit_rescale(self, image: Image.Image, merge_size: int = 2) -> Image.Image:
+        try:
+            w, h = image.size
+        except:
+            raise ValueError(str((type(image), image)))
+        patch_size = self.patch_size
+
+        if (w // patch_size) * (h // patch_size) > self.in_token_limit:
+            scale = math.sqrt(
+                self.in_token_limit / ((w // patch_size) * (h // patch_size))
+            )
+            new_w, new_h = int(w * scale), int(h * scale)
+
+            image = image.resize((new_w, new_h), Image.Resampling.BICUBIC)
+        if self.pad_input:
+            new_w, new_h = image.size
+            pad_size_h = merge_size * patch_size
+            pad_size_w = merge_size * patch_size
+
+            pad_h = (pad_size_h - new_h % pad_size_h) % pad_size_h
+            pad_w = (pad_size_w - new_w % pad_size_w) % pad_size_w
+
+            image = TF.pad(image, (0, 0, pad_w, pad_h))
+        else:
+            new_w, new_h = image.size
+            new_w = new_w - new_w % patch_size
+            new_h = new_h - new_h % patch_size
+
+            new_w = adjust_size(new_w, patch_size)
+            new_h = adjust_size(new_h, patch_size)
+
+            image = TF.center_crop(image, (new_h, new_w))
+
+        w, h = image.size
+        if w // patch_size >= 512 or h // patch_size >= 512:
+            new_h = min(patch_size * 510, h)
+            new_w = min(patch_size * 510, w)
+            image = TF.center_crop(image, (new_h, new_w))
+            # raise ValueError("Exceed pos emb")
+        return image
+
+    def _preprocess(
+        self,
+        images: Union[ImageInput, VideoInput],
+        do_resize: bool = None,
+        resample: PILImageResampling = None,
+        do_rescale: bool = None,
+        rescale_factor: float = None,
+        do_normalize: bool = None,
+        image_mean: Optional[Union[float, List[float]]] = None,
+        image_std: Optional[Union[float, List[float]]] = None,
+        do_convert_rgb: bool = None,
+        data_format: Optional[ChannelDimension] = ChannelDimension.FIRST,
+        input_data_format: Optional[Union[str, ChannelDimension]] = 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`.
+            vision_info (`List[Dict]`, *optional*):
+                Optional list of dictionaries containing additional information about vision inputs.
+            do_resize (`bool`, *optional*, defaults to `self.do_resize`):
+                Whether to resize the image.
+            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.
+            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.
+        """
+        images = make_list_of_images(images)
+
+        if input_data_format is None:
+            # We assume that all images have the same channel dimension format.
+            input_data_format = ChannelDimension.LAST if isinstance(images[0], Image.Image) else infer_channel_dimension_format(images[0])
+
+        if do_convert_rgb:
+            images = [convert_to_rgb(image) for image in images]
+
+        # All transformations expect numpy arrays.
+        images = [to_numpy_array(image) for image in images]
+
+        if is_scaled_image(images[0]) and do_rescale:
+            logger.warning_once(
+                "It looks like you are trying to rescale already rescaled images. If the input"
+                " images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again."
+            )
+
+        height, width = get_image_size(images[0], channel_dim=input_data_format)
+        resized_height, resized_width = height, width
+        processed_images = []
+
+        for image in images:
+            if do_resize:
+                resized_height, resized_width = smart_resize(
+                    height,
+                    width,
+                    factor=self.patch_size * self.merge_size,
+                    min_pixels=self.min_pixels,
+                    max_pixels=self.max_pixels,
+                )
+                image = resize(
+                    image,
+                    size=(resized_height, resized_width),
+                    resample=resample,
+                    input_data_format=input_data_format,
+                )
+
+            if do_rescale:
+                image = self.rescale(
+                    image, scale=rescale_factor, input_data_format=input_data_format
+                )
+
+            if do_normalize:
+                image = self.normalize(
+                    image=image,
+                    mean=image_mean,
+                    std=image_std,
+                    input_data_format=input_data_format,
+                )
+            image = to_channel_dimension_format(
+                image, data_format, input_channel_dim=input_data_format
+            )
+            processed_images.append(image)
+
+        patches = np.array(processed_images)
+        if data_format == ChannelDimension.LAST:
+            patches = patches.transpose(0, 3, 1, 2)
+        if patches.shape[0] == 1:
+            patches = np.tile(patches, (self.temporal_patch_size, 1, 1, 1))
+        init_patches = patches
+        channel = patches.shape[1]
+        grid_t = patches.shape[0] // self.temporal_patch_size
+        grid_h, grid_w = (
+            resized_height // self.patch_size,
+            resized_width // self.patch_size,
+        )
+        patches = patches.reshape(
+            grid_t,
+            self.temporal_patch_size,
+            channel,
+            grid_h,
+            self.patch_size,
+            grid_w,
+            self.patch_size,
+        )
+        patches = patches.transpose(0, 3, 5, 2, 1, 4, 6)
+        assert self.temporal_patch_size == 1
+        flatten_patches = patches.reshape(
+            grid_t * grid_h * grid_w, channel, self.patch_size, self.patch_size
+        )
+        return flatten_patches, (grid_t, grid_h, grid_w)
+
+    def preprocess(
+        self,
+        images: ImageInput,
+        videos: VideoInput = None,
+        do_resize: bool = None,
+        size: Dict[str, int] = None,
+        resample: PILImageResampling = None,
+        do_rescale: bool = None,
+        rescale_factor: float = None,
+        do_normalize: bool = None,
+        image_mean: Optional[Union[float, List[float]]] = None,
+        image_std: Optional[Union[float, List[float]]] = None,
+        do_convert_rgb: bool = None,
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        data_format: Optional[ChannelDimension] = ChannelDimension.FIRST,
+        input_data_format: Optional[Union[str, ChannelDimension]] = 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`.
+            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.
+
+        """
+        do_resize = do_resize if do_resize is not None else self.do_resize
+        size = size if size is not None else self.size
+        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
+        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_batched_images(images)
+        if videos is not None:
+            videos = make_batched_videos(videos)
+
+        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,
+        )
+
+        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,
+                    resample=resample,
+                    do_rescale=do_rescale,
+                    rescale_factor=rescale_factor,
+                    do_normalize=do_normalize,
+                    image_mean=image_mean,
+                    image_std=image_std,
+                    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 = {"pixel_values": pixel_values, "image_grid_thw": vision_grid_thws}
+
+        if videos is not None:
+            pixel_values, vision_grid_thws = [], []
+            for images in videos:
+                patches, video_grid_thw = self._preprocess(
+                    images,
+                    do_resize=do_resize,
+                    resample=resample,
+                    do_rescale=do_rescale,
+                    rescale_factor=rescale_factor,
+                    do_normalize=do_normalize,
+                    image_mean=image_mean,
+                    image_std=image_std,
+                    data_format=data_format,
+                    do_convert_rgb=do_convert_rgb,
+                    input_data_format=input_data_format,
+                )
+                pixel_values.extend(patches)
+                vision_grid_thws.append(video_grid_thw)
+            pixel_values = np.array(pixel_values)
+            vision_grid_thws = np.array(vision_grid_thws)
+            data = {
+                "pixel_values_videos": pixel_values,
+                "video_grid_thw": vision_grid_thws,
+            }
+
+        return BatchFeature(data=data, tensor_type=return_tensors)

inference.yml

@@ -0,0 +1,2 @@
+Global:
+  model_name: PaddleOCR-VL-0.9B

processing_paddleocr_vl.py

@@ -0,0 +1,293 @@
+# Copyright (c) 2025 PaddlePaddle Authors. 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 typing import List, Union
+import numpy as np
+import torch
+from transformers.feature_extraction_utils import BatchFeature
+from transformers.processing_utils import (
+    ProcessingKwargs,
+    ProcessorMixin,
+    Unpack,
+    VideosKwargs,
+)
+from transformers.tokenization_utils_base import PreTokenizedInput, TextInput
+
+
+ImageInput = Union[
+    "PIL.Image.Image",
+    np.ndarray,
+    "torch.Tensor",
+    List["PIL.Image.Image"],
+    List[np.ndarray],
+    List["torch.Tensor"],
+]  # noqa
+
+
+VideoInput = Union[
+    List["PIL.Image.Image"],
+    "np.ndarray",
+    "torch.Tensor",
+    List["np.ndarray"],
+    List["torch.Tensor"],
+    List[List["PIL.Image.Image"]],
+    List[List["np.ndarrray"]],
+    List[List["torch.Tensor"]],
+]  # noqa
+
+
+class PaddleOCRVLVideosProcessorKwargs(VideosKwargs, total=False):
+    fps: Union[List[float], float]
+
+
+class PaddleOCRVLProcessorKwargs(ProcessingKwargs, total=False):
+    videos_kwargs: PaddleOCRVLVideosProcessorKwargs
+    _defaults = {
+        "text_kwargs": {
+            "padding": False,
+        },
+        "videos_kwargs": {"fps": 2.0},
+    }
+
+
+class PaddleOCRVLProcessor(ProcessorMixin):
+    r"""
+    [`PaddleOCRVLProcessor`] offers all the functionalities of [`SiglipImageProcessor`] and [`Qwen2TokenizerFast`]. See the
+    [`~PaddleOCRVLProcessor.__call__`] and [`~PaddleOCRVLProcessor.decode`] for more information.
+    Args:
+        image_processor ([`SiglipImageProcessor`], *optional*):
+            The image processor is a required input.
+        tokenizer ([`Qwen2TokenizerFast`], *optional*):
+            The tokenizer is a required input.
+        chat_template (`str`, *optional*): A Jinja template which will be used to convert lists of messages
+            in a chat into a tokenizable string.
+    """
+
+    attributes = ["image_processor", "tokenizer"]
+    valid_kwargs = [
+        "chat_template",
+        "image_std",
+        "min_pixels",
+        "image_mean",
+        "merge_size",
+        "image_processor_type",
+        "temporal_patch_size",
+        "patch_size",
+        "max_pixels",
+    ]
+
+    image_processor_class = "AutoImageProcessor"
+    tokenizer_class = "AutoTokenizer"
+
+    def __init__(
+        self, image_processor=None, tokenizer=None, chat_template=None, **kwargs
+    ):
+        self.image_token = (
+            "<|IMAGE_PLACEHOLDER|>"
+            if not hasattr(tokenizer, "image_token")
+            else tokenizer.image_token
+        )
+        self.video_token = (
+            "<|video_pad|>"
+            if not hasattr(tokenizer, "video_token")
+            else tokenizer.video_token
+        )
+        super().__init__(image_processor, tokenizer, chat_template=chat_template)
+
+    def __call__(
+        self,
+        images: ImageInput = None,
+        text: Union[
+            TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]
+        ] = None,
+        videos: VideoInput = None,
+        **kwargs: Unpack[PaddleOCRVLProcessorKwargs],
+    ) -> BatchFeature:
+        """
+        Main method to prepare for the model one or several sequences(s) and image(s). This method forwards the `text`
+        and `kwargs` arguments to Qwen2TokenizerFast's [`~Qwen2TokenizerFast.__call__`] if `text` is not `None` to encode
+        the text. To prepare the vision inputs, this method forwards the `vision_infos` and `kwrags` arguments to
+        SiglipImageProcessor's [`~SiglipImageProcessor.__call__`] if `vision_infos` is not `None`.
+
+        Args:
+            images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
+                The image or batch of images to be prepared. Each image can be a PIL image, NumPy array or PyTorch
+                tensor. Both channels-first and channels-last formats are supported.
+            text (`str`, `List[str]`, `List[List[str]]`):
+                The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings
+                (pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set
+                `is_split_into_words=True` (to lift the ambiguity with a batch of sequences).
+            videos (`np.ndarray`, `torch.Tensor`, `List[np.ndarray]`, `List[torch.Tensor]`):
+                The image or batch of videos to be prepared. Each video can be a 4D NumPy array or PyTorch
+                tensor, or a nested list of 3D frames. Both channels-first and channels-last formats are supported.
+            return_tensors (`str` or [`~utils.TensorType`], *optional*):
+                If set, will return tensors of a particular framework. Acceptable values are:
+                - `'tf'`: Return TensorFlow `tf.constant` objects.
+                - `'pt'`: Return PyTorch `torch.Tensor` objects.
+                - `'np'`: Return NumPy `np.ndarray` objects.
+                - `'jax'`: Return JAX `jnp.ndarray` objects.
+
+        Returns:
+            [`BatchFeature`]: A [`BatchFeature`] with the following fields:
+
+            - **input_ids** -- List of token ids to be fed to a model. Returned when `text` is not `None`.
+            - **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when
+              `return_attention_mask=True` or if *"attention_mask"* is in `self.model_input_names` and if `text` is not
+              `None`).
+            - **pixel_values** -- Pixel values to be fed to a model. Returned when `images` is not `None`.
+            - **pixel_values_videos** -- Pixel values of videos to be fed to a model. Returned when `videos` is not `None`.
+            - **image_grid_thw** -- List of image 3D grid in LLM. Returned when `images` is not `None`.
+            - **video_grid_thw** -- List of video 3D grid in LLM. Returned when `videos` is not `None`.
+            - **second_per_grid_ts** -- List of video seconds per time grid. Returned when `videos` is not `None`.
+        """
+        output_kwargs = self._merge_kwargs(
+            PaddleOCRVLProcessorKwargs,
+            tokenizer_init_kwargs=self.tokenizer.init_kwargs,
+            **kwargs,
+        )
+
+        if images is not None:
+            image_inputs = self.image_processor(images=images, return_tensors="pt")
+            image_inputs["pixel_values"] = image_inputs["pixel_values"]
+            image_grid_thw = image_inputs["image_grid_thw"]
+
+        else:
+            image_inputs = {}
+            image_grid_thw = None
+
+        if videos is not None:
+            # TODO: add video processing
+            videos_inputs = self.image_processor(
+                images=None, videos=videos, **output_kwargs["images_kwargs"]
+            )
+            video_grid_thw = videos_inputs["video_grid_thw"]
+
+            fps = output_kwargs["videos_kwargs"].pop("fps", 2.0)
+            if isinstance(fps, (int, float)):
+                second_per_grid_ts = [
+                    self.image_processor.temporal_patch_size / fps
+                ] * len(video_grid_thw)
+            elif hasattr(fps, "__len__") and len(fps) == len(video_grid_thw):
+                second_per_grid_ts = [
+                    self.image_processor.temporal_patch_size / tmp for tmp in fps
+                ]
+            else:
+                raise ValueError(
+                    f"The length of fps ({len(fps) if hasattr(fps, '__len__') else fps}) must be equal to the length of video_grid_thw ({len(video_grid_thw)}) or fps should be a single number."
+                )
+            videos_inputs.update(
+                {"second_per_grid_ts": torch.tensor(second_per_grid_ts)}
+            )
+
+        else:
+            videos_inputs = {}
+            video_grid_thw = None
+
+        if not isinstance(text, list):
+            text = [text]
+
+        if image_grid_thw is not None:
+            index = 0
+            for i in range(len(text)):
+                while self.image_token in text[i]:
+                    text[i] = text[i].replace(
+                        self.image_token,
+                        "<|placeholder|>"
+                        * (
+                            image_grid_thw[index].prod()
+                            // self.image_processor.merge_size
+                            // self.image_processor.merge_size
+                        ),
+                        1,
+                    )
+                    index += 1
+                text[i] = text[i].replace("<|placeholder|>", self.image_token)
+
+        if video_grid_thw is not None:
+            index = 0
+            for i in range(len(text)):
+                while self.video_token in text[i]:
+                    text[i] = text[i].replace(
+                        self.video_token,
+                        "<|placeholder|>"
+                        * (
+                            video_grid_thw[index].prod()
+                            // self.image_processor.merge_size
+                            // self.image_processor.merge_size
+                        ),
+                        1,
+                    )
+                    index += 1
+                text[i] = text[i].replace("<|placeholder|>", self.video_token)
+
+        text_inputs = self.tokenizer(text, **output_kwargs["text_kwargs"])
+
+        return BatchFeature(data={**text_inputs, **image_inputs, **videos_inputs})
+
+    def batch_decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to Qwen2TokenizerFast's [`~PreTrainedTokenizer.batch_decode`]. Please
+        refer to the docstring of this method for more information.
+        """
+        return self.tokenizer.batch_decode(*args, **kwargs)
+
+    def decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to Qwen2TokenizerFast's [`~PreTrainedTokenizer.decode`]. Please refer to
+        the docstring of this method for more information.
+        """
+        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,
+    ):
+        """
+        Post-process the output of the model to decode the text.
+
+        Args:
+            generated_outputs (`torch.Tensor` or `np.ndarray`):
+                The output of the model `generate` function. The output is expected to be a tensor of shape `(batch_size, sequence_length)`
+                or `(sequence_length,)`.
+            skip_special_tokens (`bool`, *optional*, defaults to `True`):
+                Whether or not to remove special tokens in the output. Argument passed to the tokenizer's `batch_decode` method.
+            Clean_up_tokenization_spaces (`bool`, *optional*, defaults to `False`):
+                Whether or not to clean up the tokenization spaces. Argument passed to the tokenizer's `batch_decode` method.
+            **kwargs:
+                Additional arguments to be passed to the tokenizer's `batch_decode method`.
+
+        Returns:
+            `List[str]`: The decoded text.
+        """
+        return self.tokenizer.batch_decode(
+            generated_outputs,
+            skip_special_tokens=skip_special_tokens,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+
+    @property
+    def model_input_names(self):
+        tokenizer_input_names = self.tokenizer.model_input_names
+        image_processor_input_names = self.image_processor.model_input_names
+        names_from_processor = list(
+            dict.fromkeys(tokenizer_input_names + image_processor_input_names)
+        )
+        return names_from_processor + ["second_per_grid_ts"]
+
+
+__all__ = ["PaddleOCRVLProcessor", "PaddleOCRVLProcessor"]

processor_config.json

@@ -0,0 +1,6 @@
+{
+  "auto_map": {
+    "AutoProcessor": "processing_paddleocr_vl.PaddleOCRVLProcessor"
+  },
+  "processor_class": "PaddleOCRVLProcessor"
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f90f04fd8e5eb6dfa380f37d10c87392de8438dccb6768a2486b5a96ee76dba6
+size 11187679