fixed merged conflicts

.gitattributes

@@ -1,2 +0,0 @@
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text

.github/workflows/build_and_publish_to_pypi.yml

@@ -13,8 +13,6 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@master
-      with:
-        lfs: true
     - name: Set up Python 3.8
       uses: actions/setup-python@v3
       with:

.github/workflows/tests.yml

@@ -8,6 +8,7 @@ on:
     branches: [ "main" ]
   pull_request:
     branches: [ "main" ]
+  workflow_dispatch:
 
 jobs:
   unit_tests_linux:
@@ -18,8 +19,6 @@ jobs:
 
     steps:
     - uses: actions/checkout@v3
-      with:
-        lfs: true
     - name: Set up Python ${{ matrix.python-version }}
       uses: actions/setup-python@v3
       with:
@@ -42,8 +41,6 @@ jobs:
 
     steps:
     - uses: actions/checkout@v3
-      with:
-        lfs: true
     - name: Set up Python ${{ matrix.python-version }}
       uses: actions/setup-python@v3
       with:

MANIFEST.in

@@ -1,2 +0,0 @@
-recursive-include openwakeword *.onnx
-recursive-include openwakeword *.tflite

README.md

@@ -41,16 +41,20 @@ Many thanks to [TeaPoly](https://github.com/TeaPoly/speexdsp-ns-python) for thei
 
 # Usage
 
-For quick local testing, clone this repository and use the included [example script](examples/detect_from_microphone.py) to try streaming detection from a local microphone. **Important note!** The model files are stored in this repo using [git-lfs](https://git-lfs.com/); make sure it is installed on your system and if needed use `git-lfs fetch --all` to make sure the the models download correctly.
+For quick local testing, clone this repository and use the included [example script](examples/detect_from_microphone.py) to try streaming detection from a local microphone. You can individually download pre-trained models from current and past [releases](https://github.com/dscripka/openWakeWord/releases/), or you can download them using Python (see below).
 
 Adding openWakeWord to your own Python code requires just a few lines:
 
 ```python
+import openwakeword
 from openwakeword.model import Model
 
-# Instantiate the model
+# One-time download of all pre-trained models (or only select models)
+openwakeword.utils.download_models()
+
+# Instantiate the model(s)
 model = Model(
-    wakeword_models=["path/to/model.onnx"],  # can also leave this argument empty to load all of the included pre-trained models
+    wakeword_models=["path/to/model.tflite"],  # can also leave this argument empty to load all of the included pre-trained models
 )
 
 # Get audio data containing 16-bit 16khz PCM audio data from a file, microphone, network stream, etc.

openwakeword/__init__.py

@@ -5,24 +5,48 @@ from openwakeword.custom_verifier_model import train_custom_verifier
 
 __all__ = ['Model', 'VAD', 'train_custom_verifier']
 
-models = {
+FEATURE_MODELS = {
+    "embedding": {
+        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/embedding_model.tflite"),
+        "download_url": "https://github.com/dscripka/openWakeWord/releases/download/v0.5.1/embedding_model.tflite"
+    },
+    "melspectrogram": {
+        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/melspectrogram.tflite"),
+        "download_url": "https://github.com/dscripka/openWakeWord/releases/download/v0.5.1/melspectrogram.tflite"
+    }
+}
+
+VAD_MODELS = {
+    "silero_vad": {
+        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/silero_vad.onnx"),
+        "download_url": "https://github.com/dscripka/openWakeWord/releases/download/v0.5.1/silero_vad.onnx"
+    }
+}
+
+MODELS = {
     "alexa": {
-        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/alexa_v0.1.tflite")
+        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/alexa_v0.1.tflite"),
+        "download_url": "https://github.com/dscripka/openWakeWord/releases/download/v0.5.1/alexa_v0.1.tflite"
     },
     "hey_mycroft": {
-        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/hey_mycroft_v0.1.tflite")
+        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/hey_mycroft_v0.1.tflite"),
+        "download_url": "https://github.com/dscripka/openWakeWord/releases/download/v0.5.1/hey_mycroft_v0.1.tflite"
     },
     "hey_jarvis": {
-        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/hey_jarvis_v0.1.tflite")
+        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/hey_jarvis_v0.1.tflite"),
+        "download_url": "https://github.com/dscripka/openWakeWord/releases/download/v0.5.1/hey_jarvis_v0.1.tflite"
     },
     "hey_rhasspy": {
-        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/hey_rhasspy_v0.1.tflite")
+        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/hey_rhasspy_v0.1.tflite"),
+        "download_url": "https://github.com/dscripka/openWakeWord/releases/download/v0.5.1/hey_rhasspy_v0.1.tflite"
     },
     "timer": {
-        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/timer_v0.1.tflite")
+        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/timer_v0.1.tflite"),
+        "download_url": "https://github.com/dscripka/openWakeWord/releases/download/v0.5.1/timer_v0.1.tflite"
     },
     "weather": {
-        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/weather_v0.1.tflite")
+        "model_path": os.path.join(os.path.dirname(os.path.abspath(__file__)), "resources/models/weather_v0.1.tflite"),
+        "download_url": "https://github.com/dscripka/openWakeWord/releases/download/v0.5.1/weather_v0.1.tflite"
     }
 }
 
@@ -40,6 +64,6 @@ model_class_mappings = {
 
 def get_pretrained_model_paths(inference_framework="tflite"):
     if inference_framework == "tflite":
-        return [models[i]["model_path"] for i in models.keys()]
+        return [MODELS[i]["model_path"] for i in MODELS.keys()]
     elif inference_framework == "onnx":
-        return [models[i]["model_path"].replace(".tflite", ".onnx") for i in models.keys()]
+        return [MODELS[i]["model_path"].replace(".tflite", ".onnx") for i in MODELS.keys()]

openwakeword/model.py

@@ -67,7 +67,7 @@ class Model():
                                    with VAD scores above the threshold will be returned. The default value (0),
                                    disables voice activity detection entirely.
             custom_verifier_models (dict): A dictionary of paths to custom verifier models, where
-                                           the keys are the model names (corresponding to the openwakeword.models
+                                           the keys are the model names (corresponding to the openwakeword.MODELS
                                            attribute) and the values are the filepaths of the
                                            custom verifier models.
             custom_verifier_threshold (float): The score threshold to use a custom verifier model. If the score
@@ -85,7 +85,7 @@ class Model():
         wakeword_model_names = []
         if wakeword_models == []:
             wakeword_models = pretrained_model_paths
-            wakeword_model_names = list(openwakeword.models.keys())
+            wakeword_model_names = list(openwakeword.MODELS.keys())
         elif len(wakeword_models) >= 1:
             for ndx, i in enumerate(wakeword_models):
                 if os.path.exists(i):

openwakeword/resources/models/alexa_v0.1.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:6ff566a01d12670e8d9e3c59da32651db1575d17272a601b7f8a39283dfbae3e
-size 854246

openwakeword/resources/models/alexa_v0.1.tflite

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:7333a317a790070a7f3432b81d9439c779481cc4ebd67c73da7174ea3cf48397
-size 855312

openwakeword/resources/models/embedding_model.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:70d164290c1d095d1d4ee149bc5e00543250a7316b59f31d056cff7bd3075c1f
-size 1326578

openwakeword/resources/models/embedding_model.tflite

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:c0aea21eb84a4ce90a08c870da41b7a7173b45269e6a3207c71d67c40f3a59d8
-size 1330312

openwakeword/resources/models/hey_jarvis_v0.1.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:94a13cfe60075b132f6a472e7e462e8123ee70861bc3fb58434a73712ee0d2cb
-size 1271370

openwakeword/resources/models/hey_jarvis_v0.1.tflite

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:14bff778604985e1b5c19f0f7bbe477a69cf281d8db34b232b3b972411f710e2
-size 1278912

openwakeword/resources/models/hey_mycroft_v0.1.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:c2a311e8fa1338de89c31b3b46dc4dffd4af2f9a8d6ddead48893c2d301b1f18
-size 857691

openwakeword/resources/models/hey_mycroft_v0.1.tflite

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:bf9e43136afd3ca323698820a6e32a47f885ef4c30a3b8b577ec71688a9d64d8
-size 860300

openwakeword/resources/models/hey_rhasspy_v0.1.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5a9b3ed3be2910e35780e097905aa9f35a9c10038df47914cf2b3ec4d670f6ea
-size 204081

openwakeword/resources/models/hey_rhasspy_v0.1.tflite

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:01d2526b45068f565aa3849d6ec2b7abae099154fc1b496f9ef20de9ef241fe9
-size 416140

openwakeword/resources/models/melspectrogram.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ba2b0e0f8b7b875369a2c89cb13360ff53bac436f2895cced9f479fa65eb176f
-size 1087958

openwakeword/resources/models/melspectrogram.tflite

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:96fa0adccb6e8cf95cb14465409a1a2898ee4a96a85bb9ed3c7eb0e68bf163e8
-size 1092516

openwakeword/resources/models/silero_vad.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:a35ebf52fd3ce5f1469b2a36158dba761bc47b973ea3382b3186ca15b1f5af28
-size 1807522

openwakeword/resources/models/timer_v0.1.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:371e44535470a29248b3b8f1bbbbaf2525c86417fd8f75c67fcf02ae0b9626df
-size 1742475

openwakeword/resources/models/timer_v0.1.tflite

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:21d5b0267e97df64870b7aca312e2043ebed248d365698926a115a3694ff9626
-size 1743316

openwakeword/resources/models/weather_v0.1.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:8441da8e746899e8d969528d5bad5651cdd563079c05962788f77753041f60e7
-size 1149158

openwakeword/resources/models/weather_v0.1.tflite

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4178991c7aeb76670f5a56559eb4129a6f3ae6207886db8bd8094fea7d362c3f
-size 1150224

openwakeword/utils.py

@@ -25,6 +25,7 @@ from tqdm import tqdm
 import openwakeword
 from numpy.lib.format import open_memmap
 from typing import Union, List, Callable, Deque
+import requests
 
 
 # Base class for computing audio features using Google's speech_embedding
@@ -590,6 +591,79 @@ def compute_features_from_generator(generator, n_total, clip_duration, output_fi
     trim_mmap(output_file)
 
 
+# Function to download files from a URL with a progress bar
+def download_file(url, target_directory, file_size=None):
+    """A simple function to download a file from a URL with a progress bar using only the requests library"""
+    local_filename = url.split('/')[-1]
+
+    with requests.get(url, stream=True) as r:
+        if file_size is not None:
+            progress_bar = tqdm(total=file_size, unit='iB', unit_scale=True, desc=f"{local_filename}")
+        else:
+            total_size = int(r.headers.get('content-length', 0))
+            progress_bar = tqdm(total=total_size, unit='iB', unit_scale=True, desc=f"{local_filename}")
+
+        with open(os.path.join(target_directory, local_filename), 'wb') as f:
+            for chunk in r.iter_content(chunk_size=8192):
+                f.write(chunk)
+                progress_bar.update(len(chunk))
+
+    progress_bar.close()
+
+
+# Function to download models from GitHub release assets
+def download_models(
+        model_names: List[str] = [],
+        target_directory: str = os.path.join(pathlib.Path(__file__).parent.resolve(), "resources", "models")
+        ):
+    """
+    Download the specified models from the release assets in the openWakeWord GitHub repository.
+    Uses the official urls in the MODELS dictionary in openwakeword/__init__.py.
+
+    Args:
+        model_names (List[str]): The names of the models to download (e.g., hey_jarvis_v0.1). Both ONNX and
+                                 tflite models will be downloaded. If not provided (the default),
+                                 the latest versions of all models will be downloaded.
+        target_directory (str): The directory to save the models to. Defaults to the install location
+                                of openWakeWord (i.e., the `resources/models` directory).
+    Returns:
+        None
+    """
+    if not isinstance(model_names, list):
+        raise ValueError("The model_names argument must be a list of strings")
+
+    # Always download melspectrogram and embedding models, if they don't already exist
+    if not os.path.exists(target_directory):
+        os.makedirs(target_directory)
+    for feature_model in openwakeword.FEATURE_MODELS.values():
+        if not os.path.exists(os.path.join(target_directory, feature_model["download_url"].split("/")[-1])):
+            download_file(feature_model["download_url"], target_directory)
+            download_file(feature_model["download_url"].replace(".tflite", ".onnx"), target_directory)
+
+    # Always download VAD models, if they don't already exist
+    for vad_model in openwakeword.VAD_MODELS.values():
+        if not os.path.exists(os.path.join(target_directory, vad_model["download_url"].split("/")[-1])):
+            download_file(vad_model["download_url"], target_directory)
+
+    # Get all model urls
+    official_model_urls = [i["download_url"] for i in openwakeword.MODELS.values()]
+    official_model_names = [i["download_url"].split("/")[-1] for i in openwakeword.MODELS.values()]
+
+    if model_names != []:
+        for model_name in model_names:
+            url = [i for i, j in zip(official_model_urls, official_model_names) if model_name in j]
+            if url != []:
+                if not os.path.exists(os.path.join(target_directory, url[0].split("/")[-1])):
+                    download_file(url[0], target_directory)
+                    download_file(url[0].replace(".tflite", ".onnx"), target_directory)
+    else:
+        print(official_model_urls)
+        for official_model_url in official_model_urls:
+            if not os.path.exists(os.path.join(target_directory, official_model_url.split("/")[-1])):
+                download_file(official_model_url, target_directory)
+                download_file(official_model_url.replace(".tflite", ".onnx"), target_directory)
+
+
 # Handle deprecated arguments and naming (thanks to https://stackoverflow.com/a/74564394)
 def re_arg(kwarg_map):
     def decorator(func):

setup.py

@@ -32,7 +32,8 @@ setuptools.setup(
         'tflite-runtime>=2.8.0,<3; platform_system == "Linux"',
         'tqdm>=4.0,<5.0',
         'scipy>=1.3,<2',
-        'scikit-learn>=1,<2'
+        'scikit-learn>=1,<2',
+        'requests>=2.0,<3',
     ],
     extras_require={
         'test': [
@@ -44,7 +45,8 @@ setuptools.setup(
                     'types-requests',
                     'types-PyYAML',
                     'mock>=5.1,<6',
-                    'types-mock>=5.1,<6'
+                    'types-mock>=5.1,<6',
+                    'types-requests>=2.0,<3'
                 ],
         'full': [
                     'mutagen>=1.46.0,<2',

tests/test_custom_verifier_model.py

@@ -34,6 +34,9 @@ import scipy.io.wavfile
 import tempfile
 import pytest
 
+# Download models needed for tests
+openwakeword.utils.download_models(model_names=["alexa_v0.1", "hey_mycroft_v0.1"])
+
 
 # Tests
 class TestModels:

tests/test_models.py

@@ -40,6 +40,9 @@ import pickle
 import tempfile
 import mock
 
+# Download models needed for tests
+openwakeword.utils.download_models()
+
 
 # Tests
 class TestModels: