torch-audiomentations

This project publishes no release notes. Between v0.11.2 and v0.12.0 there were 4 commits, 2 of them substantive:

• Update Zenodo badge
• chore: remove librosa dependency in favor of torchaudio

This project publishes no release notes. Between v0.11.1 and v0.11.2 there were 17 commits, 13 of them substantive:

• Update Zenodo badge
• fixed bug in transform_parameters when training on multi-gpus
• added failing test case
• fix add colored noise to prevent issue.
• Bump pytest dev dependencies
• Bump numpy and scipy in dev environment
• Bump python and pip in dev environment
• Convert unittest code to pytest in test_colored_noise.py
• Fix test
• Upgrade python in CI
• Reformat code with updated black
• Convert to str for torchaudio.info call
• Backwards-compatible type-hinting

This project publishes no release notes. Between v0.11.0 and v0.11.1 there were 31 commits, 24 of them substantive:

• add SpliceOut to torch_audiomentations init
• Update Zenodo badge
• Rename spliceout to splice_out
• Include Mix, Padding and SpliceOut in the user-facing API to prepare for v0.12.0
• fix: remove unused import
• specify device for RandomCrop output variable
• Add support for constant cutoff frequency in LowPassFilter and HighPassfilter
• Update test_requirements.txt
• Update README.md
• Update README.md
• Update known issues section in readme
• Add support for min_f_decay==max_f_decay in AddColoredNoise
• Bump python version in formatting CI job
• Bump black
• Fix formatting in random_crop.py
• Reformat .flake8
• Fix typo link -> lint
• Update build status badge in readme
• Update reference from master to main
• Update readme

…and 4 more.

This project publishes no release notes. Between v0.10.1 and v0.11.0 there were 114 commits, 97 of them substantive:

• wip: add support for targets as discussed in #3
• add find_audio_files_in_paths to file utils
• replace uses of find_audio_files by find_audio_files_in_paths
• feat: add support for batches in Audio.rms_normalize
• fix: fix missing targets handling
• feat: add support for batches in Audio.rms_normalize
• wip: add Mix transform
• fix: honor "mix_target" option
• Update Zenodo badge
• fix: fix typo
• fix: add Mix to the mix
• fix: fix composition support for target
• Upgrade python, pytorch and cuda. Remove unused ffmpeg-python.
• add find_audio_files_in_paths unit test
• fix: fix two corner cases
• random crop
• init baseclass
• type conversion
• test random crop
• input sample checks

…and 77 more.

This project publishes no release notes. Between v0.10.0 and v0.10.1 there were 17 commits, 11 of them substantive:

• Update io.py
• Update io.py
• Update packaging.md
• Allow min=max snr in AddBackgroundNoise
• Test min equals max in AddBackgroundNoise
• Test bg noise length compatibility
• Update broken Zenodo badge in readme
• Fix grammar
• Remove unused audiomentations dependency
• Add support for librosa 0.9.0
• Allow newer numpy version in dev and test env

This project publishes no release notes. Between v0.9.1 and v0.10.0 there were 20 commits, 12 of them substantive:

• Implements asteroid-team/torch-audiomentations#93: Band reject filter
• Implements asteroid-team/torch-audiomentations#86: Audio time Inversion
• Band stop filer - sanity unit test
• Implements asteroid-team/torch-audiomentations#26: SomeOf and OneOf
• Updated docstring of SomeOf
• Update readme and demo with the new BandStopFilter and TimeInversion
• Reformat some code to make the linter happy
• Fixes asteroid-team/torch-audiomentations#76: Store ir_paths in transform_parameters
• Reformat code to make the linter happy
• Update docstrings
• Fix double bandwidth bug in BPF and update defaults
• Guard against too small min_bandwidth_fraction

This project publishes no release notes. Between v0.9.0 and v0.9.1 there were 7 commits, 3 of them substantive:

• Use python 3.8 in CI
• bump python version to 3.9
• Remove upper python version restriction

This project publishes no release notes. Between v0.8.0 and v0.9.0 there were 70 commits, 57 of them substantive:

• Remove support for torchaudio<=0.6
• Upgrade default python version from 3.6 to 3.7 in Github CI job
• create pitch_shift file
• added .vscode
• add torch_pitch_shift to deps
• yea no it ain't workin yet
• use new torch_pitch_shift build
• bump
• might need to bring back factorization
• boom baby, PitchShift finally works!
• ignore x == 1
• Allow for tensors with slightly different dims
• padded the wrong dimension
• fix memory leak
• fix memory leak
• writing code at 2am is clearly not very efficient
• use torch-pitch-shift instead of torch_pitch_shift
• feat: add fix_propagation_delay option
• added per_channel and per_batch (default) modes
• max_transpose_ratio needs some floaties

…and 37 more.

This project publishes no release notes. Between v0.7.0 and v0.8.0 there were 17 commits, 15 of them substantive:

• Update readme
• Require pytorch>=1.7.0
• Deprecated support for torchaudio<=0.6
• Simplify dev setup
• Implement a high-pass filter. Unfortunately it's very slow.
• Implement a low-pass filter
• Speed up the high pass filter by using julius instead of torchaudio
• Tweak the defaults for the high pass filter
• Add note on deprecated torchaudio 0.6 in setup.py
• Rephrase TODO on perf improvement
• Take cutoff calculation out of for loop
• Check the differentiability of LPF and HPF
• Fix CUDA compatibility issue in HighPassFilter and LowPassFilter
• Mention HighPassFilter and LowPassFilter in readme
• Blackify code