Skip to content
Snippets Groups Projects
  1. Feb 11, 2025
  2. Jan 25, 2025
  3. Jan 22, 2025
  4. Dec 24, 2024
  5. Nov 17, 2024
  6. Oct 26, 2024
  7. Apr 04, 2024
  8. Mar 30, 2024
  9. Mar 08, 2024
  10. Mar 07, 2024
  11. Feb 11, 2024
  12. May 27, 2023
  13. May 26, 2023
  14. Apr 06, 2023
  15. Mar 04, 2023
  16. Jan 21, 2023
  17. Jan 05, 2023
  18. Dec 07, 2022
  19. Nov 24, 2022
  20. Nov 23, 2022
  21. Oct 27, 2022
  22. Oct 26, 2022
  23. Sep 30, 2022
  24. Sep 28, 2022
  25. Aug 30, 2022
  26. Aug 29, 2022
  27. Aug 23, 2022
    • Chih-Hsuan Yen's avatar
      & various improvements · 425059fc
      Chih-Hsuan Yen authored
      * List myself as a maintainer
      * Add the link to upstream changelog, which is introduced in this version
      * Add python-numpy to as a dependency to make namcap happy
      * Drop a merged patch
      2.0.7-1
      425059fc
  28. Jun 06, 2022
    • Chih-Hsuan Yen's avatar
      backport a fix for tensorflow lite compatibility · d312b720
      Chih-Hsuan Yen authored
      Apparently, tensorflow uses flatbuffers 1.x to generate Python codes
      from flatbuffers schemes, and it's broken with flatbuffers 2.0.6. For
      example, a code snippet from a tensorflow lite tutorial [1] fails with:
      
      Traceback (most recent call last):
        File "/home/yen/tmp/t.py", line 15, in <module>
          tflite_model = converter.convert()
        File "/usr/lib/python3.10/site-packages/tensorflow/lite/python/lite.py", line 929, in wrapper
          return self._convert_and_export_metrics(convert_func, *args, **kwargs)
        File "/usr/lib/python3.10/site-packages/tensorflow/lite/python/lite.py", line 921, in _convert_and_export_metrics
          return flatbuffer_utils.convert_object_to_bytearray(model_object)
        File "/usr/lib/python3.10/site-packages/tensorflow/lite/tools/flatbuffer_utils.py", line 84, in convert_object_to_bytearray
          model_offset = model_object.Pack(builder)
        File "/usr/lib/python3.10/site-packages/tensorflow/lite/python/schema_py_generated.py", line 5951, in Pack
          operatorCodes = builder.EndVector(len(self.operatorCodes))
      TypeError: Builder.EndVector() takes 1 positional argument but 2 were given
      
      (There is another RuntimeError "Mismatching ReplicaContext", which
      seems unrelated and harmless.)
      
      As a side note, flatbuffers 2.x compatibility for tensorflow is tracked
      in [2].
      
      I've also re-enabled Python tests, which were disabled when switching
      to out-of-source builds [3], as well as added a comment for broken C++
      tests.
      
      [1] https://www.tensorflow.org/lite/convert#convert_a_keras_model_
      [2] https://github.com/tensorflow/tensorflow/issues/51487
      [3] https://github.com/archlinux/svntogit-community/commit/0786e4bd012a8a454f61acae4b142a20e842216a
      2.0.6-2
      d312b720
  29. Feb 22, 2022
  30. Dec 09, 2021
  31. Dec 01, 2021
  32. Nov 30, 2021
  33. May 10, 2021
  34. Nov 11, 2020
  35. Nov 09, 2020
  36. Mar 13, 2020
  37. Nov 01, 2019
  38. Oct 25, 2019
Loading