add mdx submission description

README.md

@@ -26,6 +26,8 @@ python main.py fit --config cfg/demucs.yaml
 
 ## Reproduce the winning submission
 
+### CDX Leaderboard A, submission ID 220319
+
 This section describes how to reproduce the models we used in our [winning submission](https://gitlab.aicrowd.com/yoyololicon/cdx-submissions/-/issues/90) on CDX leaderboard A.
 The submission consists of one HDemucs predicting all the targets and one BandSplitRNN predicitng the music from the mixture.
 
@@ -52,6 +54,34 @@ Briefly speaking, the HDemucs predicts the targets independently for each channe
 The stereo separated sources are made from a linear combination of these mono predictions.
 The separated music from the BandSplitRNN is enhanced by Wiener Filtering, and the final music predictions is the average from the two models.
 
+### MDX Leaderboard A (Label Noise), submission ID 220426
+
+This section describes how to reproduce the model we used in our [winning submission](https://gitlab.aicrowd.com/yoyololicon/mdx23-submissions/-/issues/76) on MDX leaderboard A.
+
+Firstly, we manually inspected the [label noise dataset](https://www.aicrowd.com/challenges/sound-demixing-challenge-2023/problems/music-demixing-track-mdx-23/dataset_files) and labeled the clean songs (no label noise).
+The labels are recorded in `data/lightning/label_noise.csv`.
+Then, a HDemucs was trained only on the clean labels with the following settings:
+
+* negative SDR as the loss function
+* Training occurs on random chunks and random stem combinations of the clean songs
+* Training batches are augmented and processed using different random effects
+* Due to all this randomization, validation is done also on the training dataset (no separate validation set)
+
+To reproduce the training:
+```commandline
+python main.py fit --config cfg/mdx_a/hdemucs.yaml --data.init_args.root /DATASET_ROOT/
+```
+Remember to place the label noise data under `/DATASET_ROOT/train/`.
+
+Other details:
+* Model is trained for ~800 epochs (approx. 2 weeks on 4 RTX A50000)
+* During the last ~200 epochs, the learning rate is reduced to 0.001, gradient accumulation is increased to 64, and the effect randomization chance is increased by a factor of 1.666 (e.g. 30% to 50% etc.)
+
+After training, please go to our [submission repository](https://gitlab.aicrowd.com/yoyololicon/mdx23-submissions/) and checkout the tag `submission-cm-acc64-4d-lr001-e1213-last`.
+Then, copy the checkpoint to `my_submission/acc64_4devices_lr0001_e1213_last.ckpt ` in the submission repository.
+After these steps, you have reproduced our submission!
+
+
 ## Structure
 
 * `aimless`: package root, which can be imported for submission.
@@ -88,11 +118,4 @@ For the value of `ACCESS_TOKEN` please refer to [#24](https://github.com/yoyolol
 
 ### Training Details
 
-* Clean songs (no label noise) are hand-labeled and recorded in `data/lightning/label_noise.csv`
-* A hybrid demux model is trained with negative SDR as the loss function
-* Training occurs on random chunks and random stem combinations of the clean songs
-* Training batches are augmented and processed using different random effects
-* Due to all this randomization, validation is done also on the training dataset (no separate validation set)
-* All details and hyperparameters are contained in `cfg/hdemucs.yaml`
-* Model is trained for ~800 epochs (approx. 2 weeks on 4 GPUs)
-* During the last ~200 epochs, the learning rate is reduced to 0.001, gradient accumulation is increased to 64, and the effect randomization chance is increased by a factor of 1.666 (e.g. 30% to 50% etc.)
+