Rename "get_embedding_model" to "load_embedding_model"

docs/tutorial.rst

@@ -85,7 +85,7 @@ By default, the corresponding model file is loaded every time this function is c
 
     import openl3
     import soundfile as sf
-    model = openl3.models.get_embedding_model(input_repr="mel256", content_type="music", embedding_size=6144)
+    model = openl3.models.load_embedding_model(input_repr="mel256", content_type="music", embedding_size=6144)
     emb, ts = openl3.get_embedding(audio, sr, model=model)
 
 To compute embeddings for an audio file and save them locally, you can use code like the following:
@@ -113,7 +113,7 @@ Like before, you can also load the model before processing the file so that load
     import openl3
     import numpy as np
 
-    model = openl3.models.get_embedding_model(input_repr="mel256", content_type="music", embedding_size=6144)
+    model = openl3.models.load_embedding_model(input_repr="mel256", content_type="music", embedding_size=6144)
 
     audio_filepath = '/path/to/file.wav'
     # Saves the file to '/path/to/file.npz'

openl3/cli.py

@@ -3,7 +3,7 @@
 import sys
 from openl3.openl3_exceptions import OpenL3Error
 from openl3 import process_file
-from openl3.models import get_embedding_model
+from openl3.models import load_embedding_model
 from argparse import ArgumentParser, RawDescriptionHelpFormatter, ArgumentTypeError
 from collections import Iterable
 from six import string_types
@@ -85,7 +85,7 @@ def run(inputs, output_dir=None, suffix=None, input_repr="mel256", content_type=
         sys.exit(-1)
 
     # Load model
-    model = get_embedding_model(input_repr, content_type, embedding_size)
+    model = load_embedding_model(input_repr, content_type, embedding_size)
 
     # Process all files in the arguments
     for filepath in file_list:

openl3/core.py

@@ -6,7 +6,7 @@
 from numbers import Real
 import warnings
 import keras
-from .models import get_embedding_model
+from .models import load_embedding_model
 from .openl3_exceptions import OpenL3Error
 from .openl3_warnings import OpenL3Warning
 
@@ -114,7 +114,7 @@ def get_embedding(audio, sr, model=None, input_repr="mel256",
 
     # Get embedding model
     if model is None:
-        model = get_embedding_model(input_repr, content_type, embedding_size)
+        model = load_embedding_model(input_repr, content_type, embedding_size)
 
     audio_len = audio.size
     frame_len = TARGET_SR

openl3/models.py

@@ -29,7 +29,7 @@
 }
 
 
-def get_embedding_model(input_repr, content_type, embedding_size):
+def load_embedding_model(input_repr, content_type, embedding_size):
     """
     Returns a model with the given characteristics. Loads the model
     if the model has not been loaded yet.
@@ -54,7 +54,7 @@ def get_embedding_model(input_repr, content_type, embedding_size):
         warnings.simplefilter("ignore")
         m = MODELS[input_repr]()
 
-    m.load_weights(get_embedding_model_path(input_repr, content_type))
+    m.load_weights(load_embedding_model_path(input_repr, content_type))
 
     # Pooling for final output embedding size
     pool_size = POOLINGS[input_repr][embedding_size]
@@ -64,7 +64,7 @@ def get_embedding_model(input_repr, content_type, embedding_size):
     return m
 
 
-def get_embedding_model_path(input_repr, content_type):
+def load_embedding_model_path(input_repr, content_type):
     """
     Returns the local path to the model weights file for the model
     with the given characteristics

tests/notebooks/generate_openl3_regression_data.ipynb

@@ -209,7 +209,7 @@
    "outputs": [],
    "source": [
     "# path to store output embeddings\n",
-    "output_dir = os.path.expanduser('~/Downloads/')"
+    "output_dir = os.path.expanduser('~/openl3_output/')"
    ]
   },
   {
@@ -221,9 +221,25 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": 4,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "WARNING:tensorflow:From /beegfs/jtc440/miniconda3/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:1204: calling reduce_max (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
+      "Instructions for updating:\n",
+      "keep_dims is deprecated, use keepdims instead\n",
+      "WARNING:tensorflow:From /beegfs/jtc440/miniconda3/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:1238: calling reduce_sum (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
+      "Instructions for updating:\n",
+      "keep_dims is deprecated, use keepdims instead\n",
+      "WARNING:tensorflow:From /beegfs/jtc440/miniconda3/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:1255: calling reduce_prod (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
+      "Instructions for updating:\n",
+      "keep_dims is deprecated, use keepdims instead\n"
+     ]
+    }
+   ],
    "source": [
     "# compute mel256/music/6144 regression embedding\n",
     "suffix=None\n",
@@ -247,7 +263,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 5,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -352,9 +368,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python [conda env:openl3]",
+   "display_name": "Python 3",
    "language": "python",
-   "name": "conda-env-openl3-py"
+   "name": "python3"
   },
   "language_info": {
    "codemirror_mode": {
@@ -366,7 +382,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.6.8"
+   "version": "3.6.4"
   }
  },
  "nbformat": 4,

tests/test_core.py

@@ -114,7 +114,7 @@ def test_get_embedding():
     assert not np.any(np.isnan(emb1))
 
     # Make sure we can load a model and pass it in
-    model = openl3.models.get_embedding_model("linear", "env", 6144)
+    model = openl3.models.load_embedding_model("linear", "env", 6144)
     emb1load, ts1load = openl3.get_embedding(audio, sr,
         model=model, center=True, hop_size=hop_size, verbose=1)
     assert np.all(np.abs(emb1load - emb1) < tol)

tests/test_models.py

@@ -1,59 +1,59 @@
-from openl3.models import get_embedding_model, get_embedding_model_path
+from openl3.models import load_embedding_model, load_embedding_model_path
 
 
-def test_get_embedding_model_path():
-    embedding_model_path = get_embedding_model_path('linear', 'music')
+def test_load_embedding_model_path():
+    embedding_model_path = load_embedding_model_path('linear', 'music')
     assert '/'.join(embedding_model_path.split('/')[-2:]) == 'openl3/openl3_audio_linear_music.h5'
 
-    embedding_model_path = get_embedding_model_path('linear', 'env')
+    embedding_model_path = load_embedding_model_path('linear', 'env')
     assert '/'.join(embedding_model_path.split('/')[-2:]) == 'openl3/openl3_audio_linear_env.h5'
 
-    embedding_model_path = get_embedding_model_path('mel128', 'music')
+    embedding_model_path = load_embedding_model_path('mel128', 'music')
     assert '/'.join(embedding_model_path.split('/')[-2:]) == 'openl3/openl3_audio_mel128_music.h5'
 
-    embedding_model_path = get_embedding_model_path('mel128', 'env')
+    embedding_model_path = load_embedding_model_path('mel128', 'env')
     assert '/'.join(embedding_model_path.split('/')[-2:]) == 'openl3/openl3_audio_mel128_env.h5'
 
-    embedding_model_path = get_embedding_model_path('mel256', 'music')
+    embedding_model_path = load_embedding_model_path('mel256', 'music')
     assert '/'.join(embedding_model_path.split('/')[-2:]) == 'openl3/openl3_audio_mel256_music.h5'
 
-    embedding_model_path = get_embedding_model_path('mel256', 'env')
+    embedding_model_path = load_embedding_model_path('mel256', 'env')
     assert '/'.join(embedding_model_path.split('/')[-2:]) == 'openl3/openl3_audio_mel256_env.h5'
 
 
-def test_get_embedding_model():
-    m = get_embedding_model('linear', 'music', 6144)
+def test_load_embedding_model():
+    m = load_embedding_model('linear', 'music', 6144)
     assert m.output_shape[1] == 6144
 
-    m = get_embedding_model('linear', 'music', 512)
+    m = load_embedding_model('linear', 'music', 512)
     assert m.output_shape[1] == 512
 
-    m = get_embedding_model('linear', 'env', 6144)
+    m = load_embedding_model('linear', 'env', 6144)
     assert m.output_shape[1] == 6144
 
-    m = get_embedding_model('linear', 'env', 512)
+    m = load_embedding_model('linear', 'env', 512)
     assert m.output_shape[1] == 512
 
-    m = get_embedding_model('mel128', 'music', 6144)
+    m = load_embedding_model('mel128', 'music', 6144)
     assert m.output_shape[1] == 6144
 
-    m = get_embedding_model('mel128', 'music', 512)
+    m = load_embedding_model('mel128', 'music', 512)
     assert m.output_shape[1] == 512
 
-    m = get_embedding_model('mel128', 'env', 6144)
+    m = load_embedding_model('mel128', 'env', 6144)
     assert m.output_shape[1] == 6144
 
-    m = get_embedding_model('mel128', 'env', 512)
+    m = load_embedding_model('mel128', 'env', 512)
     assert m.output_shape[1] == 512
 
-    m = get_embedding_model('mel256', 'music', 6144)
+    m = load_embedding_model('mel256', 'music', 6144)
     assert m.output_shape[1] == 6144
 
-    m = get_embedding_model('mel256', 'music', 512)
+    m = load_embedding_model('mel256', 'music', 512)
     assert m.output_shape[1] == 512
 
-    m = get_embedding_model('mel256', 'env', 6144)
+    m = load_embedding_model('mel256', 'env', 6144)
     assert m.output_shape[1] == 6144
 
-    m = get_embedding_model('mel256', 'env', 512)
+    m = load_embedding_model('mel256', 'env', 512)
     assert m.output_shape[1] == 512