Upload folder using huggingface_hub

config.gin

@@ -0,0 +1,99 @@
+# Parameters for AudioDataModule:
+# ==============================================================================
+AudioDataModule.num_workers = 20
+
+# Parameters for AudioDataset:
+# ==============================================================================
+AudioDataset.half_precision = True
+AudioDataset.mono = True
+AudioDataset.new_freq = 16000
+AudioDataset.num_frames = 480000
+AudioDataset.orig_freq = 16000
+
+# Parameters for build_dev_datamodule:
+# ==============================================================================
+build_dev_datamodule.datamodule = @discotube
+
+# Parameters for build_module:
+# ==============================================================================
+build_module.ckpt_path = 'model.ckpt'
+build_module.module = @modules.maskingmodel.MaskingModel
+build_module.net = @nets.conformer.Conformer
+build_module.representation = @nets.melspectrogram.MelSpectrogram
+
+# Parameters for Conformer:
+# ==============================================================================
+Conformer.alpha_deepnorm = 2.6321480259049848
+Conformer.beta_deepnorm = 0.022386873579657126
+Conformer.conv_kernel_size = 5
+Conformer.depth = 24
+Conformer.dropout = 0.2
+Conformer.embed_dim = 1024
+Conformer.input_dropout = 0.0
+Conformer.mlp_ratio = 4.0
+Conformer.mlp_residual_factor = 4.0
+Conformer.num_heads = 8
+Conformer.num_patches = 460
+Conformer.patch_size = (96, 4)
+Conformer.use_deepnorm = True
+Conformer.use_rope = True
+
+# Parameters for CosineAnnealingCallback:
+# ==============================================================================
+CosineAnnealingCallback.eta_min = 1e-07
+CosineAnnealingCallback.warmup_steps = 30000
+
+# Parameters for DiscotubeAudioDataModule:
+# ==============================================================================
+DiscotubeAudioDataModule.batch_size = 32
+DiscotubeAudioDataModule.data_dir = ''
+DiscotubeAudioDataModule.filelist_train = ''
+DiscotubeAudioDataModule.filelist_val = ''
+
+# Parameters for MaskingModel:
+# ==============================================================================
+MaskingModel.codebook_dim = 16
+MaskingModel.codebook_size = 8192
+MaskingModel.diff_input = False
+MaskingModel.lr = 0.0001
+MaskingModel.mask_prob = 0.6
+MaskingModel.mask_seconds = 0.4
+MaskingModel.num_codebooks = 4
+MaskingModel.plot_tokens = False
+MaskingModel.seed = 0
+MaskingModel.weight_decay = 0.01
+
+# Parameters for MelSpectrogram:
+# ==============================================================================
+MelSpectrogram.freq_mask_param = 0
+MelSpectrogram.hop_len = 256
+MelSpectrogram.mel_scale = 'slaney'
+MelSpectrogram.n_mel = 96
+MelSpectrogram.norm = 'slaney'
+MelSpectrogram.norm_mean = 2.06755686098554
+MelSpectrogram.norm_std = 1.268292820667291
+MelSpectrogram.power = 2
+MelSpectrogram.sr = 16000
+MelSpectrogram.stretch_factor = 1
+MelSpectrogram.time_mask_param = 0
+MelSpectrogram.win_len = 512
+MelSpectrogram.patch_size = (96, 4)
+
+# Parameters for train:
+# ==============================================================================
+train.params = \
+    {'accelerator': 'gpu',
+     'devices': 4,
+     'log_every_n_steps': 50,
+     'max_steps': 400000,
+     'num_nodes': 1,
+     'num_sanity_val_steps': 0,
+     'precision': 'bf16-mixed',
+     'strategy': 'ddp_find_unused_parameters_true'}
+train.wandb_params = \
+    {'entity': 'mtg-upf',
+     'group': 'masking_conformer',
+     'name': 'mask_conformer_rope_multi4_large',
+     'offline': True,
+     'project': 'mtg-ssl',
+     'save_dir': '/gpfs/projects/upf97/logs/'}

model.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04384614600f75333e13c9ff950ceed1cf0fd296c44f7ce82aa32e2b4512e99a
+size 7374069142