Update torchaudio to 0.9 for tensorflow-gpu compatibility

[jbitton]

Summary:
As called out in #28, there are some conflicting dependencies between torchaudio/torch 0.8.1/1.8.1 and tensorflow-gpu.

However, as discovered in pytorch/audio#1595, upgrading to v0.9 etc actually resolve this issue.

Thus, I update the torchaudio/torch versions in our requirements.txt and I also updated our numpy requirement so there are no conflicting dependencies between tf-gpu and augly :)

I verified on my side that all unit tests still pass and that setup.py finishes as expected with no errors. I also update setup.py to add our README to our PyPI page.

Ran:

python -m unittest augly.tests.audio_tests.functional_unit_tests
----------------------------------------------------------------------
Ran 21 tests in 4.683s

OK

python -m unittest augly.tests.audio_tests.transforms_unit_tests
----------------------------------------------------------------------
Ran 24 tests in 1.892s

OK

python -m unittest augly.tests.image_tests.functional_unit_tests
----------------------------------------------------------------------
Ran 32 tests in 32.442s

OK (skipped=2)

python -m unittest augly.tests.image_tests.transforms_unit_tests
----------------------------------------------------------------------
Ran 39 tests in 12.182s

OK (skipped=3)

python -m unittest augly.tests.video_tests.transforms.cv2_tests
----------------------------------------------------------------------
Ran 5 tests in 72.686s

OK

and

python setup.py install
.....
Using /home/jbitton/anaconda3/envs/testing_reqs/lib/python3.9/site-packages
Finished processing dependencies for augly==0.1.1

Differential Revision: D29292956

[facebook-github-bot]

This pull request was exported from Phabricator. Differential Revision: D29292956

[jbitton]

@zpapakipos let me know if running this on your side works without issue. if yes, let's merge + update the PyPI package!

[zpapakipos]

Tested on my side.

Installed using new reqs locally:

cd AugLy/
git pull
git checkout change_reqs
conda create -n augly_change_reqs -y
conda activate augly_change_reqs
cd ..
pip install -e AugLy/

Ran unit tests:

python -m unittest augly.tests.audio_tests.functional_unit_tests
----------------------------------------------------------------------
Ran 21 tests in 2.664s

OK

python -m unittest augly.tests.audio_tests.transforms_unit_tests
----------------------------------------------------------------------
Ran 24 tests in 1.069s

OK

python -m unittest augly.tests.image_tests.functional_unit_tests
----------------------------------------------------------------------
Ran 32 tests in 27.815s

OK (skipped=2)

python -m unittest augly.tests.image_tests.transforms_unit_tests
----------------------------------------------------------------------
Ran 39 tests in 10.127s

OK (skipped=3)

python -m unittest augly.tests.text_tests.functional_unit_tests
----------------------------------------------------------------------
Ran 12 tests in 0.059s

OK

python -m unittest augly.tests.text_tests.transforms_unit_tests
----------------------------------------------------------------------
Ran 12 tests in 0.016s

OK

python -m unittest augly.tests.video_tests.transforms.composite_tests
----------------------------------------------------------------------
Ran 11 tests in 31.113s

OK

Built & uploaded new pypi package:

# <changed version number in setup.py to 0.1.2>
git add setup.py
git commit -m "Update version # to 0.1.2"
git push
conda activate pypi
python -m build --wheel
twine upload dist/*

Tested new pypi package (in colab audio example notebook):

!pip install -U augly
!sudo apt-get install python3-magic

import os
import augly.audio as audaugs
import augly.utils as utils
from IPython.display import display, Audio

# Get input audio
input_audio = os.path.join(
    utils.TEST_URI, "audio", "inputs", "vad-go-stereo-44100.wav"
)
display(Audio(input_audio))

from augly.audio.utils import validate_and_load_audio

input_audio_arr, sr = validate_and_load_audio(input_audio)
meta = []
aug = audaugs.PeakingEqualizer()
aug_audio, sr = aug(input_audio_arr, sample_rate=sr, metadata=meta)
display(Audio(aug_audio, rate=sr))
meta

[zpapakipos]

Looks great, and all my testing works! Thanks for making this change :) I just updated the version # in setup.py since I changed that in the change_reqs branch before uploading the package.