Update model configuration and dataset parameters

recipes/intel_ndns/spiking_fullsubnet/full320_sub224_Pk32-128_1-1-1_D5-3-1.toml

@@ -0,0 +1,93 @@
+[meta]
+save_dir = "exp"
+description = "Train a model using Generative Adversarial Networks (GANs)"
+seed = 20220815
+
+[trainer]
+path = "trainer.Trainer"
+[trainer.args]
+debug = false
+max_steps = 0
+max_epochs = 200
+max_grad_norm = 10
+save_max_score = true
+save_ckpt_interval = 1
+max_patience = 20
+plot_norm = true
+validation_interval = 1
+max_num_checkpoints = 20
+scheduler_name = "constant_schedule_with_warmup"
+warmup_steps = 0
+warmup_ratio = 0.00
+gradient_accumulation_steps = 1
+
+[loss_function]
+path = "torch.nn.MSELoss"
+[loss_function.args]
+
+[optimizer]
+path = "torch.optim.AdamW"
+[optimizer.args]
+lr = 1e-3
+
+[model]
+path = "audiozen.models.spiking_fullsubnet.modeling_spiking_fullsubnet.SpikingFullSubNet"
+[model.args]
+n_fft = 512
+hop_length = 128
+win_length = 512
+fdrc = 0.5
+fb_input_size = 64
+fb_hidden_size = 320
+fb_num_layers = 2
+fb_proj_size = 64
+fb_output_activate_function = false
+sb_hidden_size = 224
+sb_num_layers = 2
+freq_cutoffs = [0, 32, 128, 256]
+df_orders = [5, 3, 1]
+center_freq_sizes = [1, 1, 1]
+neighbor_freq_sizes = [15, 15, 15]
+use_pre_layer_norm_fb = true
+use_pre_layer_norm_sb = true
+bn = true
+shared_weights = true
+sequence_model = "GSN"
+num_spks = 1
+
+[acoustics]
+n_fft = 512
+hop_length = 128
+win_length = 512
+sr = 16000
+
+[train_dataset]
+path = "dataloader.DNSAudio"
+[train_dataset.args]
+root = "/datasets/datasets_fullband/training_set/"
+limit = false
+offset = 0
+
+[train_dataset.dataloader]
+batch_size = 6
+num_workers = 6
+drop_last = true
+pin_memory = true
+
+[validate_dataset]
+path = "dataloader.DNSAudio"
+[validate_dataset.args]
+root = "/datasets/datasets_fullband/validation_set/"
+train = false
+[validate_dataset.dataloader]
+batch_size = 16
+num_workers = 8
+
+[test_dataset]
+path = "dataloader.DNSAudio"
+[test_dataset.args]
+root = "/nfs/xhao/data/intel_ndns/test_set/"
+train = false
+[test_dataset.dataloader]
+batch_size = 1
+num_workers = 0

recipes/intel_ndns/spiking_fullsubnet/full320_sub224_Pk32-128_2-32-64_D5-3-1.toml

@@ -0,0 +1,93 @@
+[meta]
+save_dir = "exp"
+description = "Train a model using Generative Adversarial Networks (GANs)"
+seed = 20220815
+
+[trainer]
+path = "trainer.Trainer"
+[trainer.args]
+debug = false
+max_steps = 0
+max_epochs = 200
+max_grad_norm = 10
+save_max_score = true
+save_ckpt_interval = 5
+max_patience = 20
+plot_norm = true
+validation_interval = 5
+max_num_checkpoints = 20
+scheduler_name = "constant_schedule_with_warmup"
+warmup_steps = 0
+warmup_ratio = 0.00
+gradient_accumulation_steps = 1
+
+[loss_function]
+path = "torch.nn.MSELoss"
+[loss_function.args]
+
+[optimizer]
+path = "torch.optim.AdamW"
+[optimizer.args]
+lr = 1e-3
+
+[model]
+path = "audiozen.models.spiking_fullsubnet.modeling_spiking_fullsubnet.SpikingFullSubNet"
+[model.args]
+n_fft = 512
+hop_length = 128
+win_length = 512
+fdrc = 0.5
+fb_input_size = 64
+fb_hidden_size = 320
+fb_num_layers = 2
+fb_proj_size = 64
+fb_output_activate_function = false
+sb_hidden_size = 224
+sb_num_layers = 2
+freq_cutoffs = [0, 32, 128, 256]
+df_orders = [5, 3, 1]
+center_freq_sizes = [2, 32, 64]
+neighbor_freq_sizes = [15, 15, 15]
+use_pre_layer_norm_fb = true
+use_pre_layer_norm_sb = true
+bn = true
+shared_weights = true
+sequence_model = "GSN"
+num_spks = 1
+
+[acoustics]
+n_fft = 512
+hop_length = 128
+win_length = 512
+sr = 16000
+
+[train_dataset]
+path = "dataloader.DNSAudio"
+[train_dataset.args]
+root = "/datasets/datasets_fullband/training_set/"
+limit = false
+offset = 0
+
+[train_dataset.dataloader]
+batch_size = 32
+num_workers = 8
+drop_last = true
+pin_memory = true
+
+[validate_dataset]
+path = "dataloader.DNSAudio"
+[validate_dataset.args]
+root = "/datasets/datasets_fullband/validation_set/"
+train = false
+[validate_dataset.dataloader]
+batch_size = 1
+num_workers = 0
+
+[test_dataset]
+path = "dataloader.DNSAudio"
+[test_dataset.args]
+root = "/nfs/xhao/data/intel_ndns/test_set/"
+train = false
+[test_dataset.dataloader]
+batch_size = 1
+num_workers = 0

recipes/intel_ndns/spiking_fullsubnet/model.py

@@ -431,21 +431,22 @@ def forward(self, input):
         win_length=512,
         fdrc=0.5,
         fb_input_size=64,
-        fb_hidden_size=256,
+        fb_hidden_size=320,
         fb_num_layers=2,
         fb_proj_size=64,
         fb_output_activate_function=None,
-        sb_hidden_size=128,
+        sb_hidden_size=224,
         sb_num_layers=2,
-        freq_cutoffs=[0, 20, 80, 256],
-        df_orders=[2, 2, 2],
-        center_freq_sizes=[2, 10, 22],
-        neighbor_freq_sizes=[8, 16, 32],
+        freq_cutoffs=[0, 32, 128, 256],
+        df_orders=[5, 3, 1],
+        center_freq_sizes=[2, 32, 64],
+        neighbor_freq_sizes=[15, 15, 15],
         use_pre_layer_norm_fb=True,
         use_pre_layer_norm_sb=True,
-        bn=False,
-        shared_weights=False,
+        bn=True,
+        shared_weights=True,
         sequence_model="GSN",
+        num_spks=1,
     )
 
     input = torch.rand(2, 16000)