Error running pip install -r requirements.txt in sparrow-data/ocr

[pcaselas]

Here's the swig version that I have:

SWIG Version 4.2.1

Compiled with clang++ [aarch64-apple-darwin23.2.0]

Configured options: +pcre

And here's the error message:

      fitz/fitz_wrap.c:29582:59: warning: incompatible pointer types passing 'struct Font *' to parameter of type 'struct Font *' [-Wincompatible-pointer-types]
          result = (PyObject *)TextWriter_append(arg1,arg2,arg3,arg4,arg5,arg6,arg7,arg8);
                                                                ^~~~
      fitz/fitz_wrap.c:17285:102: note: passing argument to parameter 'font' here
      SWIGINTERN PyObject *TextWriter_append(struct TextWriter *self,PyObject *pos,char *text,struct Font *font,float fontsize,char *language,int right_to_left,int small_caps){
                                                                                                           ^
      fitz/fitz_wrap.c:29736:15: warning: incompatible pointer types passing 'struct Font *' to parameter of type 'struct Font *' [-Wincompatible-pointer-types]
        delete_Font(arg1);
                    ^~~~
      fitz/fitz_wrap.c:17355:42: note: passing argument to parameter 'self' here
      SWIGINTERN void delete_Font(struct Font *self){
                                               ^
      74 warnings and 12 errors generated.
      error: command '/usr/bin/clang' failed with exit code 1
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
error: legacy-install-failure

× Encountered error while trying to install package.
╰─> PyMuPDF

[abaranovskis-redsamurai]

I'm not using SWIG

[pcaselas]

Before installing Swig I was getting this error:

      PyMuPDF/setup.py: Finished building mupdf.
      PyMuPDF/setup.py: sys.platform='darwin'
      PyMuPDF/setup.py: library_dirs=['mupdf-1.20.3-source/build/release', 'mupdf-1.20.3-source/build/release']
      PyMuPDF/setup.py: libraries=['mupdf', 'mupdf-third']
      PyMuPDF/setup.py: include_dirs=['mupdf-1.20.3-source/include', 'mupdf-1.20.3-source/include/mupdf', 'mupdf-1.20.3-source/thirdparty/freetype/include']
      PyMuPDF/setup.py: extra_link_args=[]
      running install
      running build
      running build_py
      running build_ext
      building 'fitz._fitz' extension
      swigging fitz/fitz.i to fitz/fitz_wrap.c
      swig -python -o fitz/fitz_wrap.c fitz/fitz.i
      error: command 'swig' failed: No such file or directory
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
error: legacy-install-failure

× Encountered error while trying to install package.
╰─> PyMuPDF

note: This is an issue with the package mentioned above, not pip.
hint: See above for output from the failure.

That's why I did brew install swig

[abaranovskis-redsamurai]

For me it works based on instructions from README.

[pcaselas]

Are you running it on an ARM Mac?

Seems like it does install if I upgrade paddleocr to 2.7.3, however:

 python api.py

A module that was compiled using NumPy 1.x cannot be run in
NumPy 2.0.0 as it may crash. To support both 1.x and 2.x
versions of NumPy, modules must be compiled with NumPy 2.0.
Some module may need to rebuild instead e.g. with 'pybind11>=2.12'.

If you are a user of the module, the easiest solution will be to
downgrade to 'numpy<2' or try to upgrade the affected module.
We expect that some modules will need time to support NumPy 2.

[abaranovskis-redsamurai]

I'm using Intel Mac.

Numpy error - this is Python dependency issue. Solve it by specifying lower Numpy version in requirements.txt:

numpy==1.26.4

[pcaselas]

That solves the dependencies problem, however when I run

 curl -X 'POST' \
  'http://127.0.0.1:8001/api/v1/sparrow-ocr/inference' \
  -H 'accept: application/json' \
  -H 'Content-Type: multipart/form-data' \
  -F 'file=' \
  -F 'image_url=https://raw.githubusercontent.com/katanaml/sparrow/main/sparrow-ml/llm/data/inout-20211211_001.jpg'

it seems to enter an infinite loop in here...result = model.ocr(bytes_data, cls=True) and never responds Tried uploading the file instead of image_url and the same thing happened.

It seems it's not compatible with Apple M series chips.

[abaranovskis-redsamurai]

Yes, I think you are right about M chips. Here is the discussion, related to PaddleOCR and M chips: https://github.com/PaddlePaddle/PaddleOCR/issues/11706

[abaranovskis-redsamurai]

I think the easiest solution to wrap this service into Docker container.