InvalidArgumentError: scores has incompatible shape [Op:CombinedNonMaxSuppression]

[laumecha]

I am trying to perform the inference of one image using a quantized model using the "detect.py" file. I have quantized the original yolov4 model using QKeras and then, I have transform it to SaveModel format. However, now I am getting the error:

Traceback (most recent call last):
  File "/mnt/beegfs/gap/laumecha/conda-qkeras/tensorflow-yolov4-tflite/detect.py", line 134, in <module>
    app.run(main)
  File "/mnt/beegfs/gap/laumecha/miniconda3/envs/qkeras_env/lib/python3.9/site-packages/absl/app.py", line 312, in run
    _run_main(main, args)
  File "/mnt/beegfs/gap/laumecha/miniconda3/envs/qkeras_env/lib/python3.9/site-packages/absl/app.py", line 258, in _run_main
    sys.exit(main(argv))
  File "/mnt/beegfs/gap/laumecha/conda-qkeras/tensorflow-yolov4-tflite/detect.py", line 113, in main
    boxes, scores, classes, valid_detections = tf.image.combined_non_max_suppression(
  File "/mnt/beegfs/gap/laumecha/miniconda3/envs/qkeras_env/lib/python3.9/site-packages/tensorflow/python/util/dispatch.py", line 206, in wrapper
    return target(*args, **kwargs)
  File "/mnt/beegfs/gap/laumecha/miniconda3/envs/qkeras_env/lib/python3.9/site-packages/tensorflow/python/ops/image_ops_impl.py", line 5101, in combined_non_max_suppression
    return gen_image_ops.combined_non_max_suppression(
  File "/mnt/beegfs/gap/laumecha/miniconda3/envs/qkeras_env/lib/python3.9/site-packages/tensorflow/python/ops/gen_image_ops.py", line 358, in combined_non_max_suppression
    _ops.raise_from_not_ok_status(e, name)
  File "/mnt/beegfs/gap/laumecha/miniconda3/envs/qkeras_env/lib/python3.9/site-packages/tensorflow/python/framework/ops.py", line 6897, in raise_from_not_ok_status
    six.raise_from(core._status_to_exception(e.code, message), None)
  File "<string>", line 3, in raise_from
tensorflow.python.framework.errors_impl.InvalidArgumentError: scores has incompatible shape [Op:CombinedNonMaxSuppression]

Before converting to SavedModel format I have seen that the topology of the non-quantized and the quantized models have the same layers formats and outputs. So I assume that the output should be the same but I do not have much experience with SavedModel models.