Qianfan-OCR-MLX-BF16

Qianfan-OCR — MLX BF16

MLX conversion of baidu/Qianfan-OCR for Apple Silicon.

Usage

from PIL import Image
from mlx_vlm import load, generate
from mlx_vlm.prompt_utils import apply_chat_template

IMAGENET_MEAN = (0.485, 0.456, 0.406)
IMAGENET_STD = (0.229, 0.224, 0.225)

def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size):
    best_ratio_diff = float("inf")
    best_ratio = (1, 1)
    area = width * height

    for ratio in target_ratios:
        target_aspect_ratio = ratio[0] / ratio[1]
        ratio_diff = abs(aspect_ratio - target_aspect_ratio)

        if ratio_diff < best_ratio_diff:
            best_ratio_diff = ratio_diff
            best_ratio = ratio
        elif ratio_diff == best_ratio_diff:
            if area > 0.5 * image_size * image_size * ratio[0] * ratio[1]:
                best_ratio = ratio

    return best_ratio


def dynamic_preprocess(image, min_num=1, max_num=12, image_size=448, use_thumbnail=False):
    orig_width, orig_height = image.size
    aspect_ratio = orig_width / orig_height

    target_ratios = set(
        (i, j)
        for n in range(min_num, max_num + 1)
        for i in range(1, n + 1)
        for j in range(1, n + 1)
        if i * j <= max_num and i * j >= min_num
    )
    target_ratios = sorted(target_ratios, key=lambda x: x[0] * x[1])

    target_aspect_ratio = find_closest_aspect_ratio(
        aspect_ratio, target_ratios, orig_width, orig_height, image_size
    )

    target_width = image_size * target_aspect_ratio[0]
    target_height = image_size * target_aspect_ratio[1]
    blocks = target_aspect_ratio[0] * target_aspect_ratio[1]

    resized_img = image.resize((target_width, target_height))
    processed_images = []

    tiles_per_row = target_width // image_size
    for i in range(blocks):
        box = (
            (i % tiles_per_row) * image_size,
            (i // tiles_per_row) * image_size,
            ((i % tiles_per_row) + 1) * image_size,
            ((i // tiles_per_row) + 1) * image_size,
        )
        split_img = resized_img.crop(box)
        processed_images.append(split_img)

    if use_thumbnail and len(processed_images) != 1:
        thumbnail_img = image.resize((image_size, image_size))
        processed_images.append(thumbnail_img)

    return processed_images


def load_image_tiles(image_file, input_size=448, max_num=12):
    image = Image.open(image_file).convert("RGB")
    return dynamic_preprocess(
        image,
        image_size=input_size,
        use_thumbnail=True,
        max_num=max_num,
    )


MODEL_PATH = "mat50013/Qianfan-OCR-MLX-BF16"
IMAGE_PATH = "./documents/pic.jpeg"

model, processor = load(MODEL_PATH)

images = load_image_tiles(IMAGE_PATH, input_size=448, max_num=12)

prompt = "Parse this document to Markdown."

formatted_prompt = apply_chat_template(
    processor,
    model.config,
    prompt,
    num_images=len(images),
)

result = generate(
    model,
    processor,
    formatted_prompt,
    image=images,
    max_tokens=16384,
    verbose=False,
)

print(result.text if hasattr(result, "text") else result)