Version 0.7.1

What's Changed

• Update link to easy colab by @mikeoliphant in #292
• Fix requirements by @sdatkinson in #295

Full Changelog: v0.7.0...v0.7.1

Version 0.7.0

What's Changed

• Remove bin/train/colab.ipynb by @sdatkinson in #274
• Add "nano" preset for WaveNet and LSTM by @mikeoliphant in #279
• Add ignore_checks kwarg explicitly to help users see what to change by @sdatkinson in #283
• Log sample rate in .nam files by @sdatkinson in #284
• Add "nano" info to easy colab by @mikeoliphant in #286

Full Changelog: v0.6.0...v0.7.0

Version 0.6.0

What's Changed

• Only update metadata flag after batch by @vossenv in #205
• New reamping file v2_0_0.wav by @sdatkinson in #217
• CPU-friendly training parameters by @sdatkinson in #219
• Strong and weak matches for recognized training files by @sdatkinson in #220
• Fix model device handling by @sdatkinson in #225
• Fix broken enumeration by @sdatkinson in #226
• Importing weights for WaveNet by @sdatkinson in #234
• Add LSTM to easy colab by @mikeoliphant in #185
• Fix misspelled key for mrstft_weight by @sdatkinson in #237
• Tweak v2_0_0.wav checks by @sdatkinson in #239
• Restrict Python to 3.10 and earlier by @renebohne in #245
• Support modeling cabs by @sdatkinson in #250
• Improve v2_0_0.wav by @sdatkinson in #253
• Rename demo output file from "overdrive.wav" to "output.wav" by @sdatkinson in #256
• Fix AudioShapeMismatchError mixed up arguments by @sdatkinson in #258
• Tweak cab modeling by @sdatkinson in #260
• Tune delay calculation threshold by @sdatkinson in #262
• Show 4 significant digits for checkpoint ESR by @mikeoliphant in #265
• Support wavio latest by @kostas-emman in #222

New Contributors

• @renebohne made their first contribution in #245
• @kostas-emman made their first contribution in #222

Full Changelog: v0.5.2...v0.6.0

Version 0.5.2

What's Changed

• Fix learning rate decay by @sdatkinson in #215

Full Changelog: v0.5.1...v0.5.2

Version 0.5.1

What's Changed

• Export model indices start at zero by @sdatkinson in #113
• Add rate param to wav_to_np by @sdatkinson in #115
• Fix graceful shutdowns for Lightning in Windows by @sdatkinson in #118
• Add architecture information to easy collab by @mikeoliphant in #116
• Easy colab typo fix by @mikeoliphant in #119
• Add Apple Metal support for faster training on M1/M2 by @phillipmself in #122
• Better error messages by @sdatkinson in #125
• Update colab.ipynb by @sdatkinson in #129
• Models compute their output level by @sdatkinson in #135
• FFT Loss by @sdatkinson in #143
• Batched training and silent by @vossenv in #145
• Added experimental multiresolution STFT loss using auraloss. by @honkkis in #147
• Issue 126 update readme turn key training by @cdicle in #150
• Auraloss dependency by @sdatkinson in #153
• Run multi-resolution STFT loss on-device unless it fails by @sdatkinson in #165
• Metadata in model files, loudness as first metadatum by @sdatkinson in #168
• Smaller .nam files by @sdatkinson in #179
• fix typo in recurrent.py by @eltociear in #183
• Update core.py by @fabprezja in #189
• Move "metadata" before "weights" in exported model by @mikeoliphant in #193
• Change button info string to be more helpful, add digit to ESR by @vossenv in #174
• Update gui.py by @fabprezja in #190
• Metadata in GUI trainer by @sdatkinson in #202
• Metadata in easy_colab.ipynb by @sdatkinson in #203
• Colab: Ensure model is on CPU before plotting prediction vs target by @sdatkinson in #209
• I'm all out of bubblegum by @vossenv in #206

New Contributors

• @mikeoliphant made their first contribution in #116
• @phillipmself made their first contribution in #122
• @vossenv made their first contribution in #145
• @honkkis made their first contribution in #147
• @cdicle made their first contribution in #150
• @eltociear made their first contribution in #183
• @fabprezja made their first contribution in #189

Full Changelog: v0.5.0...v0.5.1

Version 0.5.0

What's Changed

• Improvements to Easy Colab by @sdatkinson in #55
• New input signal for easy_colab.ipynb by @sdatkinson in #57
• Add TensorBoard monitoring to easy_colab.ipynb by @sdatkinson in #59
• Expose head_scale as a settable parameter, default to 0.02 by @sdatkinson in #61
• Update environment definition by @sdatkinson in #67
• Fix ESR loss by @sdatkinson in #73
• Utility module by @sdatkinson in #75
• Update training input defaults by @sdatkinson in #78
• Add different pre-made architectures to easy mode by @sdatkinson in #81
• Single-file model export by @sdatkinson in #77
• Turn off snapshot export by default by @sdatkinson in #79
• Non-integer delays by @sdatkinson in #84
• Bug fixes by @sdatkinson in #87
• ONNX support for LSTMs by @sdatkinson in #93
• Improve data validation by @sdatkinson in #97
• GUI Trainer by @sdatkinson in #101

Full Changelog: v0.4.0...v0.5.0

Version 0.4.0

New features

• Pre-emphasis filter and loss (#42)
• WaveNet and CatWaveNet models (#43)

Bug fixes and performance improvements

• Fix bug exporting parametric models (#41)
• Remove redundant "nx" in data set config; let the model provide it w/ its receptive field.
• Faster data indexing in ConcatDatasets with many data sets (#45)
• Miscellaneous improvements to the LSTM model.
• Improvements to the Colab notebook.

Version 0.3.0

Version 0.3

New features

• Hypernet parametric model (#23) (Not supported by plugin yet)
• DC loss (#31)
• ESR loss (#33)
• LSTM model (#35)
• Conditional LSTM model (#36)

Version 0.2.1

Mainly bug fixes and a new way to run using Colab.

New features

• Colab notebook (Issue 18)

Bug fixes

• delay=0 no longer deletes data in nam.data.Dataset (Issue 15)
• setup.py now properly installs (Issue 16)

Version 0.2.0

Version 0.2

Complete rework

• Switch from TensorFlow1 to PyTorch Lightning as the ML framework.
• Complete rework of training/running/exporting scripts.
• Linear model (i.e. an impulse response)
• Convolutional neural net model
• Exporting utility for use with https://github.com/sdatkinson/iPlug2