included prediction speed experiment in the main repo

.gitignore

@@ -168,3 +168,4 @@ datasets
 *.hdf5
 *.no
 *.npz
+*.csv

impsy/impsy.py

@@ -5,7 +5,7 @@
 from .dataset import dataset
 from .train import train
 from .interaction import run
-from .tests import test_mdrnn
+from .tests import test_mdrnn, prediction_speed_test
 
 @click.group()
 def cli():
@@ -18,5 +18,6 @@ def main():
     cli.add_command(train)
     cli.add_command(run)
     cli.add_command(test_mdrnn)
+    cli.add_command(prediction_speed_test)
     # runs the command line interface
     cli()

impsy/tests.py

@@ -1,6 +1,10 @@
 import click
 import time
 from .utils import mdrnn_config
+import pandas as pd
+
+
+pd.set_option('display.float_format', lambda x: '%.2f' % x)
 
 
 @click.command(name="test-mdrnn")
@@ -25,7 +29,6 @@ def build_network(sess, dimension, units, mixes, layers):
                                             n_hidden_units=units,
                                             n_mixtures=mixes,
                                             layers=layers)
-    print("MDRNN Loaded.")
     return net
 
   start_build = time.time()
@@ -34,3 +37,67 @@ def build_network(sess, dimension, units, mixes, layers):
       sess = tf.Session()
   build_network(sess, 4, model_config["units"], model_config["mixes"], model_config["layers"])
   print("Done. That took", time.time() - start_build, "seconds.")
+
+
+@click.command(name="test-speed")
+def prediction_speed_test():
+    """This command runs a speed test experiment with different sized MDRNN models. The output is written to a CSV file."""
+    start_import = time.time()
+    import impsy.mdrnn as mdrnn
+    import tensorflow.compat.v1 as tf
+    print("Done. That took", time.time() - start_import, "seconds.")
+
+    def build_network(sess, compute_graph, net_config):
+        """Build the MDRNN."""
+        mdrnn.MODEL_DIR = "./models/"
+        tf.keras.backend.set_session(sess)
+        with compute_graph.as_default():
+            net = mdrnn.PredictiveMusicMDRNN(mode=mdrnn.NET_MODE_RUN,
+                                                dimension=net_config['dimension'],
+                                                n_hidden_units=net_config['units'],
+                                                n_mixtures=net_config['mixes'],
+                                                layers=net_config['layers'])
+        return net
+
+    def request_rnn_prediction(input_value, net):
+        """ Accesses a single prediction from the RNN. """
+        start = time.time()
+        output_value = net.generate_touch(input_value)
+        time_delta = time.time() - start
+        return output_value, time_delta
+
+    def run_test(tests, config):
+        times = []
+        compute_graph = tf.Graph()
+        with compute_graph.as_default():
+            sess = tf.Session()
+        net = build_network(sess, compute_graph, config)
+        for i in range(tests):
+            ## Predictions.
+            item = mdrnn.random_sample(out_dim=config['dimension'])
+            tf.keras.backend.set_session(sess)
+            with compute_graph.as_default():
+                rnn_output, t = request_rnn_prediction(item, net)
+            out_dict = {
+                'time': t, 
+                'mixes': config['mixes'],
+                'layers': config['layers'],
+                'units': config['units'],
+                'dimension': config['dimension']}
+            times.append(out_dict)
+        # clean up
+        tf.keras.backend.clear_session()
+        sess.close()
+        return times
+
+    experiments = []
+    mdrnn_units = [64, 128, 256, 512]
+    dimensions = [2, 3, 4, 5, 6, 7, 8, 9]
+    for un in mdrnn_units:
+        for dim in dimensions:
+            net_config = {"mixes": 5, "layers": 2, "units": un, "dimension": dim}
+            times = run_test(100, net_config)
+            experiments.extend(times)
+    total_experiment = pd.DataFrame.from_records(experiments)
+    total_experiment.to_csv("total_exp.csv")
+    print(total_experiment.describe())