* added reverb/noisify/AudioReader/RandomChoice/RandomApply

* bug fixed

* transform name changes

* work around for bug in paddle's groupnorm

* upgraded to use float64 inside for high numerical acc

* fixed docstring, add nn.Layer as super for Noisify

* fixed docstring

* added mfcc func/trans and dct function

* updated unit test

* add dtype to control datatype in win function

* add dtype control in transforms

* add dtype control in functionals

* updated test

* added dtype control, updated test

* added sound classication

* added liscense, clean code, add pre-commit

* update req

* moved to PaddlePaddle-models

* code re-structure

* update README.md

* update README.md

* Update README.md

* add audioset training

* default resample mode to kaiser_fast

* delete some comments

* precommit check

* sha->rev

* add config.ymal

* remove SoundClassification from paddlespeech, since it's in PaddleAudio now

* add labels

* remove old labels

* update code

* empty

* #5300

* add evaluate, etc

* remove trace|

* import evaluate

* path update

* precommit check

* recover slowfast

* restore README.md to paddle:develop

* refactor

* update readme

* update README.md

* refactor

* refactor

* refactor

* refactor

* precommit fixed

* update README.md

* Update README.md

* Update README.md

* Update train.py

changed prefixed, removed some comments

* add wav file for testing

* bug fixed eval,new checkpoint map=0.416

* Update README.md

* added dcase task1b example

* update README.md

* code fixed for last review

* fixed level string formating

* fixed according to PR reviews

* added wav2vec2.0

* restore datatsets

* add liscense, remove scipy, move test_audio to cloud

* remove 3rd-party dependency:pathos

* add testing for wav2vec2

* update README.md

* updated README.md, added librispeech results

* Revert "updated README.md, added librispeech results"

This reverts commit da4012958e8e0bf2d7f4b608f74518583dd7d73b.

* code fixed from reviews

* add librispeech test

* remove pathos imports

* updated README.md

* update README.md

* minor-fix according to code reviews

* updated README_LP.MD

* fixed according to code review

* fixed according to code review

* added preprocessing example

* removed dcase2021_task1b from examples

* remove preprocessing from examples

* added amsoftmax to losses

* added eer/min_dcf to metrics

* updated __init__.py

* add stft,spectrogram, melspectrogram, log-melspectrogram

* add _internal, transoform, functional to imports

* add new module: functional

* add new module: window.py to _internel/

* add correspoding new unit-test for the new modules

* added ISTFT

* clean code and docstring, clean unit test

* clean code and docstring

* functional

* added back preprocessing

* add README.md

* remove preprocessing for now

* clean code, add doc

* change _internal to signal

* add new transoforms

* add new functionals

* add eps to amsoftmax, return the prediction

* add ffmpeg backend

* remove dithering in depth-convert, add ffmpeg to backend

* add Mudecode/enccode/RandomCodec

* changed variable name, fixed bug

* use namedtuple for returning

* refactor utils

* refactor

* add melspectrogram/spectrogram, add doc string

* add doc string, clean code

* rename window to windowing

* updated docstring, minor bug fixed

* move losses.py to future examples

* remove mu_encode/decode

* refactor

* move metrics to future examples

* remove features/

* naming changes for mu law algorithms

* update test, add testing utils

* fixed import

* fixed import

* fixed duplicate output in logging

* add code examples, shape info, etc

* add doc for public functions

* make backend controllable

* fixed coding stype in docstring

* export parameters for power_to_db LogMel transform

* default to_db to False to be consistent with functional

* fixed typo in docstring

* added sound classication

* added liscense, clean code, add pre-commit

* update req

* moved to PaddlePaddle-models

* code re-structure

* update README.md

* update README.md

* Update README.md

* add audioset training

* default resample mode to kaiser_fast

* delete some comments

* precommit check

* sha->rev

* add config.ymal

* remove SoundClassification from paddlespeech, since it's in PaddleAudio now

* add labels

* remove old labels

* update code

* empty

* #5300

* add evaluate, etc

* remove trace|

* import evaluate

* path update

* precommit check

* recover slowfast

* restore README.md to paddle:develop

* refactor

* update readme

* update README.md

* refactor

* refactor

* refactor

* refactor

* precommit fixed

* update README.md

* Update README.md

* Update README.md

* Update train.py

changed prefixed, removed some comments

* add wav file for testing

* bug fixed eval,new checkpoint map=0.416

* Update README.md

* added dcase task1b example

* update README.md

* code fixed for last review

* fixed level string formating

* fixed according to PR reviews

* added wav2vec2.0

* restore datatsets

* add liscense, remove scipy, move test_audio to cloud

* remove 3rd-party dependency:pathos

* add testing for wav2vec2

* update README.md

* updated README.md, added librispeech results

* Revert "updated README.md, added librispeech results"

This reverts commit da4012958e8e0bf2d7f4b608f74518583dd7d73b.

* code fixed from reviews

* add librispeech test

* remove pathos imports

* updated README.md

* update README.md

* minor-fix according to code reviews

* updated README_LP.MD

* fixed according to code review

* fixed according to code review

* added preprocessing example

* removed dcase2021_task1b from examples

* remove preprocessing from examples

* added amsoftmax to losses

* added eer/min_dcf to metrics

* updated __init__.py

* add stft,spectrogram, melspectrogram, log-melspectrogram

* add _internal, transoform, functional to imports

* add new module: functional

* add new module: window.py to _internel/

* add correspoding new unit-test for the new modules

* added ISTFT

* clean code and docstring, clean unit test

* clean code and docstring

* functional

* added back preprocessing

* add README.md

* remove preprocessing for now

* clean code, add doc

* change _internal to signal

* add new transoforms

* add new functionals

* add eps to amsoftmax, return the prediction

* add ffmpeg backend

* remove dithering in depth-convert, add ffmpeg to backend

* add Mudecode/enccode/RandomCodec

* changed variable name, fixed bug

* use namedtuple for returning

* refactor utils

* refactor

* add melspectrogram/spectrogram, add doc string

* add doc string, clean code

* rename window to windowing

* updated docstring, minor bug fixed

* move losses.py to future examples

* remove mu_encode/decode

* refactor

* move metrics to future examples

* remove features/

* naming changes for mu law algorithms

* update test, add testing utils

* fixed import

* fixed import

* fixed duplicate output in logging

* add code examples, shape info, etc

* add doc for public functions

* make backend controllable

* fixed coding stype in docstring

* added sound classication

* added liscense, clean code, add pre-commit

* update req

* moved to PaddlePaddle-models

* code re-structure

* update README.md

* update README.md

* Update README.md

* add audioset training

* default resample mode to kaiser_fast

* delete some comments

* precommit check

* sha->rev

* add config.ymal

* remove SoundClassification from paddlespeech, since it's in PaddleAudio now

* add labels

* remove old labels

* update code

* empty

* #5300

* add evaluate, etc

* remove trace|

* import evaluate

* path update

* precommit check

* recover slowfast

* restore README.md to paddle:develop

* refactor

* update readme

* update README.md

* refactor

* refactor

* refactor

* refactor

* precommit fixed

* update README.md

* Update README.md

* Update README.md

* Update train.py

changed prefixed, removed some comments

* add wav file for testing

* bug fixed eval,new checkpoint map=0.416

* Update README.md

* added dcase task1b example

* update README.md

* code fixed for last review

* fixed level string formating

* fixed according to PR reviews

* added wav2vec2.0

* restore datatsets

* add liscense, remove scipy, move test_audio to cloud

* remove 3rd-party dependency:pathos

* add testing for wav2vec2

* update README.md

* updated README.md, added librispeech results

* Revert "updated README.md, added librispeech results"

This reverts commit da4012958e8e0bf2d7f4b608f74518583dd7d73b.

* code fixed from reviews

* add librispeech test

* remove pathos imports

* updated README.md

* update README.md

* minor-fix according to code reviews

* updated README_LP.MD

* fixed according to code review

* fixed according to code review

* added preprocessing example

* removed dcase2021_task1b from examples

* remove preprocessing from examples

* added amsoftmax to losses

* added eer/min_dcf to metrics

* updated __init__.py

* add stft,spectrogram, melspectrogram, log-melspectrogram

* add _internal, transoform, functional to imports

* add new module: functional

* add new module: window.py to _internel/

* add correspoding new unit-test for the new modules

* added ISTFT

* clean code and docstring, clean unit test

* clean code and docstring

* functional

* added back preprocessing

* add README.md

* remove preprocessing for now

* clean code, add doc

* change _internal to signal

* add new transoforms

* add new functionals

* add eps to amsoftmax, return the prediction

* add ffmpeg backend

* remove dithering in depth-convert, add ffmpeg to backend

* add Mudecode/enccode/RandomCodec

* changed variable name, fixed bug

* use namedtuple for returning

* refactor utils

* refactor

* add melspectrogram/spectrogram, add doc string

* add doc string, clean code

* rename window to windowing

* updated docstring, minor bug fixed

* move losses.py to future examples

* remove mu_encode/decode

* refactor

* move metrics to future examples

* remove features/

* naming changes for mu law algorithms

* update test, add testing utils

* fixed import

* added sound classication

* added liscense, clean code, add pre-commit

* update req

* moved to PaddlePaddle-models

* code re-structure

* update README.md

* update README.md

* Update README.md

* add audioset training

* default resample mode to kaiser_fast

* delete some comments

* precommit check

* sha->rev

* add config.ymal

* remove SoundClassification from paddlespeech, since it's in PaddleAudio now

* add labels

* remove old labels

* update code

* empty

* #5300

* add evaluate, etc

* remove trace|

* import evaluate

* path update

* precommit check

* recover slowfast

* restore README.md to paddle:develop

* refactor

* update readme

* update README.md

* refactor

* refactor

* refactor

* refactor

* precommit fixed

* update README.md

* Update README.md

* Update README.md

* Update train.py

changed prefixed, removed some comments

* add wav file for testing

* bug fixed eval,new checkpoint map=0.416

* Update README.md

* added dcase task1b example

* update README.md

* code fixed for last review

* fixed level string formating

* fixed according to PR reviews

* added wav2vec2.0

* restore datatsets

* add liscense, remove scipy, move test_audio to cloud

* remove 3rd-party dependency:pathos

* add testing for wav2vec2

* update README.md

* updated README.md, added librispeech results

* Revert "updated README.md, added librispeech results"

This reverts commit da4012958e8e0bf2d7f4b608f74518583dd7d73b.

* code fixed from reviews

* add librispeech test

* remove pathos imports

* updated README.md

* update README.md

* minor-fix according to code reviews

* updated README_LP.MD

* fixed according to code review

* fixed according to code review

* added preprocessing example

* removed dcase2021_task1b from examples

* remove preprocessing from examples

* added amsoftmax to losses

* added eer/min_dcf to metrics

* updated __init__.py
Co-authored-by: Nranchlai <=ranchlai@163.com>

* Add aishell and librispeech dataset

* Add aishell and librispeech dataset

* Add aishell and librispeech dataset

* Add UrbanAudioVisualScenes Dataset

* Update features api

* added sound classication

* added liscense, clean code, add pre-commit

* update req

* moved to PaddlePaddle-models

* code re-structure

* update README.md

* update README.md

* Update README.md

* add audioset training

* default resample mode to kaiser_fast

* delete some comments

* precommit check

* sha->rev

* add config.ymal

* remove SoundClassification from paddlespeech, since it's in PaddleAudio now

* add labels

* remove old labels

* update code

* empty

* #5300

* add evaluate, etc

* remove trace|

* import evaluate

* path update

* precommit check

* recover slowfast

* restore README.md to paddle:develop

* refactor

* update readme

* update README.md

* refactor

* refactor

* refactor

* refactor

* precommit fixed

* update README.md

* Update README.md

* Update README.md

* Update train.py

changed prefixed, removed some comments

* add wav file for testing

* bug fixed eval,new checkpoint map=0.416

* Update README.md
Co-authored-by: Nranchlai <=ranchlai@163.com>

* Update audio tagging example

* Update audio tagging example

* Add sound classification example

* Add sound classification example

* Refactor code in paddleaudio/models

* Upgrade __getitem__ to support on the fly feature extraction