Remove umx

Fix classifier for non-panns models

Update to latest classifier inference

Restore classifier, move shell scripts to scripts

Update CSV logger

More cleanup, edit configs

Add new scripts/configs for eval and easy usage. Rename configs.

Merge branch 'main' into classifier-inference

Fix remfx all effects selection

Add shuffling effect order, all effects present for chain_inference to cfg

Update datagen silence threshold to 1e-4

moving to multi-way binary classification task

Merge branch 'cjs--classifier-v2' of https://github.com/mhrice/RemFx into classifier-inference

WIP: datapoint writing

updates to score logging fx classifier

adding mixup augmentation and fixing ordering of labels

Add classifier into inference. Comment-out hearbaseline for now

changing linear layers to MLP

Add resume from checkpoint for training

Fix loss issue in chain_inference

Add more experiments

Fix causal cropping for input metrics

Fix DCUNet

Fix input metrics

Fix custom inferencing issues

Add custom-inference dataset

Fix metric logging with order

Add custom effect ordering during chain inference

Fix circular import issue

Added intermediate logging of chain_inference

Add effects chain inference code

Fix bug in trying to import removed models

Remove previous DPTNet/DCUNet implementations

Tune effects, replace models with asteroid, remove silence

Skip chunk if too quiet

Add dsd100 dataset

Add target cropping if outputs are different length

Add TCN

Update callbacks, debug new models

Init dcunet and dptnet

Fix new dataset to work for remfx training

adding multi-label classification task with CNN

Refactor dataset to apply multiple effects at a time

Fix umx_compressor path. Only log max 10 audio items2

Fix LRScheduler and ckpt_path

Add new loss to umx

Add gradient clipping and lr scheduler

Fix not logging FAD on test input data2

Fix not logging FAD on test input data

Remove FAD logging on input data during train