Jetson Nano - PaddlePaddle-gpu installation issues

Hello! I am trying to install paddleocr + paddlepaddle-gpu on device Jetson Nano. Here are the specs of the device:

Model: NVIDIA Jetson Nano Developer Kit - Jetpack 4.6 [L4T 32.6.1]
NV Power Mode[0]: MAXN
Hardware:
- P-Number: p3448-0000
- Module: NVIDIA Jetson Nano (4 GB ram)
Platform:
- Distribution: Ubuntu 18.04 Bionic Beaver
- Release: 4.9.253-gcca4bf6
jtop:
- Version: 4.3.2
- Service: Inactive
Libraries:
- CUDA: 10.2.300
- cuDNN: 8.2.1.32
- TensorRT: 8.0.1.6
- VPI: 1.1.15
- Vulkan: 1.2.70

Installation was easy and smooth. I followed the steps to download appropriate .whl file from NVIDIA repo:
0) creating venv with python 3.7.17

  1. pip install -U https://paddle-inference-lib.bj.bcebos.com/2.5.2/python/Jetson/jetpack4.6_gcc7.5/nano/paddlepaddle_gpu-2.5.2-cp37-cp37m-linux_aarch64.whl
  2. pip install paddleocr (which sucessfully installing paddleocr=2.7.3 automatically)
  3. then running python with the following code:
import paddle
paddle.utils.run_check()

gives me the following error:

Running verify PaddlePaddle program ...
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/tpcp/.pyenv/versions/paddle_gpu_252_cp37/lib/python3.7/site-packages/paddle/utils/install_check.py", line 249, in run_check
    _run_static_single(use_cuda, use_xpu)
  File "/home/tpcp/.pyenv/versions/paddle_gpu_252_cp37/lib/python3.7/site-packages/paddle/utils/install_check.py", line 135, in _run_static_single
    out, parameter_list=[weight.name]
  File "/home/tpcp/.pyenv/versions/paddle_gpu_252_cp37/lib/python3.7/site-packages/decorator.py", line 232, in fun
    return caller(func, *(extras + args), **kw)
  File "/home/tpcp/.pyenv/versions/paddle_gpu_252_cp37/lib/python3.7/site-packages/paddle/fluid/wrapped_decorator.py", line 25, in __impl__
    return wrapped_func(*args, **kwargs)
  File "/home/tpcp/.pyenv/versions/paddle_gpu_252_cp37/lib/python3.7/site-packages/paddle/fluid/framework.py", line 473, in __impl__
    return func(*args, **kwargs)
  File "/home/tpcp/.pyenv/versions/paddle_gpu_252_cp37/lib/python3.7/site-packages/paddle/fluid/backward.py", line 2180, in append_backward
    target_grad_block, fwd_op_num, grad_to_var, grad_info_map
  File "/home/tpcp/.pyenv/versions/paddle_gpu_252_cp37/lib/python3.7/site-packages/paddle/fluid/backward.py", line 1756, in _append_backward_vars_
    op_desc.infer_shape(block.desc)
RuntimeError: (NotFound) The kernel `sum_grad` is not registered.
  [Hint: Expected iter != kernels_.end(), but received iter == kernels_.end().] (at /home/paddle/data/xly/workspace/23282/Paddle/paddle/phi/core/kernel_factory.cc:289)
  [operator < reduce_sum_grad > error]

Could you please advice and help me what’s going on there? OCR is not working for me either…
I already ensured that paddle is installed with cuda and was appropriate chosen from this link https://www.paddlepaddle.org.cn/inference/v2.5/guides/install/download_lib.html in front of row Jetpack4.6: nv_jetson-cuda10.2-trt8.0-nano.

>>> import paddle
>>> paddle.is_compiled_with_cuda()
True

Hi,

Do you follow the document below?

If not, please give it a try.
Thanks.