PP-OCRv6 small text detection MLX

This is an MLX-format conversion of PaddlePaddle/PP-OCRv6_small_det_safetensors for use with mlx-vlm.

import mlx.core as mx
from PIL import Image
from mlx_vlm import load

model, processor = load("mikoy92/PP-OCRv6-small-det-mlx")
image = Image.open("document.png")

inputs = processor(image)
outputs = model(**inputs)
mx.eval(outputs.logits)

result = processor.post_process_object_detection(
    outputs,
    target_sizes=inputs["target_sizes"],
)[0]
print(result["boxes"])