...

.gitignore

@@ -4,6 +4,7 @@ __pycache__/
 *$py.class
 
 output/
+input/
 
 # C extensions
 *.so

scripts/trainer.py

@@ -4,6 +4,11 @@
 from tensorflow.keras.layers import LSTM, Dense
 from music21 import stream, note, midi
 from datetime import datetime
+import json
+
+# Assume 'dataset' is your generated or real music dataset
+with open('settings.json') as f:
+    settings = json.load(f)
 
 # Function to generate a synthetic dataset
 def generate_dataset(num_samples=1000, sequence_length=50, num_classes=128):
@@ -24,6 +29,10 @@ def sequence_to_stream(sequence):
 def generate_music(model, seed_sequence, length=1000, temperature=1.0):
     generated_sequence = seed_sequence.copy()
 
+    # Get user input for the length and temperature
+    length = settings.get("length", input("Enter the length of the generated sequence: "))
+    temperature = settings.get("temperature", input("Enter the temperature for randomness (e.g., 0.5 for more randomness, 2.0 for less randomness): "))
+
     for _ in range(length):
         # Predict the next set of notes
         next_notes_prob = model.predict(np.expand_dims(generated_sequence, axis=0))
@@ -66,6 +75,10 @@ def build_model(input_shape, output_shape):
 dataset = generate_dataset()
 X, y = dataset['input'], dataset['output']
 
+# Get user input for the number of epochs and batch size if not provided in settings
+epochs = settings.get("epochs", int(input("Enter the number of epochs: ")))
+batch_size = settings.get("batch_size", int(input("Enter the batch size: ")))
+
 # Assume 'model' is your trained LSTM model
 # Make sure to replace 'input_shape' and 'output_shape' with the actual shapes from your dataset
 input_shape = X.shape[1:]
@@ -75,7 +88,7 @@ def build_model(input_shape, output_shape):
 model = build_model(input_shape, output_shape)
 
 # Train the model
-model.fit(X, y, epochs=50, batch_size=64)
+model.fit(X, y, epochs=epochs, batch_size=batch_size)
 
 # Save the trained model with a timestamp
 model_timestamp = datetime.now().strftime('%Y%m%d%H%M%S')
@@ -87,12 +100,16 @@ def build_model(input_shape, output_shape):
 # Example usage
 seed_sequence = X[np.random.randint(0, X.shape[0])]  # Use a random sequence from your dataset as a seed
 
+# Get user input for the length and temperature if not provided in settings
+length = settings.get("length", int(input("Enter the length of the generated sequence: ")))
+temperature = settings.get("temperature", float(input("Enter the temperature for randomness: ")))
+
 # Generate music with the loaded model
-generated_music_sequence = generate_music(loaded_model, seed_sequence, length=100)
+generated_music_sequence = generate_music(loaded_model, seed_sequence, length=length, temperature=temperature)
 
 # Convert the generated sequence to a music21 stream for visualization
 generated_music_stream = sequence_to_stream(generated_music_sequence)
 
 # Save the generated music to a MIDI file with a timestamp
 midi_filename = save_to_midi(generated_music_stream, 'output/generated_output')
-print(f'Generated music saved to: {midi_filename}')
+print(f'Generated music saved to: {midi_filename}')

settings.json

@@ -0,0 +1,6 @@
+{
+    "epochs": 50,
+    "batch_size": 64,
+    "length": 1000,
+    "temperature": 1.0
+  }