WIP polypennfcn 31ks and 3ks configs

config/polypennfcn-31ks-batch.py

@@ -0,0 +1,52 @@
+MODULE = 'penn'
+
+# Configuration name
+CONFIG = 'polypennfcn-31ks-batch'
+
+# gset only
+DATASETS = ['gset']
+
+EVALUATION_DATASETS = ['gset']
+
+STEPS = 50000
+
+LOG_INTERVAL = 500
+
+CHECKPOINT_INTERVAL = 5000  # steps
+
+# audio parameters
+SAMPLE_RATE = 11025
+
+# the original hopsize is 256 samples, this is 4 times less than that
+HOPSIZE = 64 
+
+WINDOW_SIZE = HOPSIZE
+
+# use only the voiced frames
+VOICED_ONLY = True
+
+STRING_INDEX = None
+
+# poly pitch net model
+MODEL = 'polypennfcn'
+
+PITCH_CATS = 6
+
+GSET_SPLIT_PLAYERS = True
+
+NUM_TRAINING_FRAMES = 128 
+
+BATCH_SIZE = 32 
+
+NORMALIZATION = 'batch'
+
+FCN = True
+
+DECODER = 'argmax'
+
+# Batch size to use for evaluation
+EVALUATION_BATCH_SIZE = None
+
+KERNEL_SIZE = 31
+
+PADDING_SIZE = 15

config/polypennfcn-3ks-batch.py

@@ -0,0 +1,52 @@
+MODULE = 'penn'
+
+# Configuration name
+CONFIG = 'polypennfcn-3ks-batch'
+
+# gset only
+DATASETS = ['gset']
+
+EVALUATION_DATASETS = ['gset']
+
+STEPS = 50000
+
+LOG_INTERVAL = 500
+
+CHECKPOINT_INTERVAL = 5000  # steps
+
+# audio parameters
+SAMPLE_RATE = 11025
+
+# the original hopsize is 256 samples, this is 4 times less than that
+HOPSIZE = 64 
+
+WINDOW_SIZE = HOPSIZE
+
+# use only the voiced frames
+VOICED_ONLY = True
+
+STRING_INDEX = None
+
+# poly pitch net model
+MODEL = 'polypennfcn'
+
+PITCH_CATS = 6
+
+GSET_SPLIT_PLAYERS = True
+
+NUM_TRAINING_FRAMES = 128 
+
+BATCH_SIZE = 32 
+
+NORMALIZATION = 'batch'
+
+FCN = True
+
+DECODER = 'argmax'
+
+# Batch size to use for evaluation
+EVALUATION_BATCH_SIZE = None
+
+KERNEL_SIZE = 3
+
+PADDING_SIZE = 1