add setup.py

.github/workflows/ci.yml

@@ -1,4 +1,4 @@
-name: Deep Speaker TCN CI
+name: Deep Speaker CI
 
 on: [push, pull_request]
 

cli.py

@@ -6,13 +6,13 @@
 
 import click
 
-from audio import Audio
-from batcher import KerasFormatConverter
-from constants import SAMPLE_RATE, NUM_FRAMES
-from test import test
-from train import start_training
-from utils import ClickType as Ct, ensures_dir
-from utils import init_pandas
+from deep_speaker.audio import Audio
+from deep_speaker.batcher import KerasFormatConverter
+from deep_speaker.constants import SAMPLE_RATE, NUM_FRAMES
+from deep_speaker.test import test
+from deep_speaker.train import start_training
+from deep_speaker.utils import ClickType as Ct, ensures_dir
+from deep_speaker.utils import init_pandas
 
 logger = logging.getLogger(__name__)
 

audio.py → deep_speaker/audio.py

@@ -8,8 +8,8 @@
 from python_speech_features import fbank
 from tqdm import tqdm
 
-from constants import SAMPLE_RATE, NUM_FBANKS
-from utils import find_files, ensures_dir
+from deep_speaker.constants import SAMPLE_RATE, NUM_FBANKS
+from deep_speaker.utils import find_files, ensures_dir
 
 logger = logging.getLogger(__name__)
 

batcher.py → deep_speaker/batcher.py

@@ -9,10 +9,10 @@
 import numpy as np
 from tqdm import tqdm
 
-from audio import pad_mfcc, Audio
-from constants import NUM_FRAMES, NUM_FBANKS
-from conv_models import DeepSpeakerModel
-from utils import ensures_dir, load_pickle, load_npy, train_test_sp_to_utt
+from deep_speaker.audio import pad_mfcc, Audio
+from deep_speaker.constants import NUM_FRAMES, NUM_FBANKS
+from deep_speaker.conv_models import DeepSpeakerModel
+from deep_speaker.utils import ensures_dir, load_pickle, load_npy, train_test_sp_to_utt
 
 logger = logging.getLogger(__name__)
 
@@ -221,7 +221,7 @@ def get_random_batch(self, batch_size, is_test=False):
         return batch_x, batch_y
 
     def get_batch_train(self, batch_size):
-        from test import batch_cosine_similarity
+        from deep_speaker.test import batch_cosine_similarity
         # s1 = time()
         self.batch_count += 1
         if self.batch_count % self.history_every == 0:
@@ -424,7 +424,7 @@ def __init__(self, kx_train, ky_train, kx_test, ky_test, model: DeepSpeakerModel
     def get_batch(self, batch_size, is_test=False, predict=None):
         if predict is None:
             predict = self.model.m.predict
-        from test import batch_cosine_similarity
+        from deep_speaker.test import batch_cosine_similarity
         num_triplets = batch_size // 3
         inputs = []
         k = 2  # do not change this.

conv_models.py → deep_speaker/conv_models.py

@@ -17,8 +17,8 @@
 from tensorflow.keras.models import Model
 from tensorflow.keras.optimizers import Adam
 
-from constants import NUM_FBANKS, SAMPLE_RATE, NUM_FRAMES
-from triplet_loss import deep_speaker_loss
+from deep_speaker.constants import NUM_FBANKS, SAMPLE_RATE, NUM_FRAMES
+from deep_speaker.triplet_loss import deep_speaker_loss
 
 logger = logging.getLogger(__name__)
 

test.py → deep_speaker/test.py

@@ -3,12 +3,12 @@
 import numpy as np
 from tqdm import tqdm
 
-from audio import Audio
-from batcher import LazyTripletBatcher
-from constants import NUM_FBANKS, NUM_FRAMES, CHECKPOINTS_TRIPLET_DIR, BATCH_SIZE
-from conv_models import DeepSpeakerModel
-from eval_metrics import evaluate
-from utils import load_best_checkpoint, enable_deterministic
+from deep_speaker.audio import Audio
+from deep_speaker.batcher import LazyTripletBatcher
+from deep_speaker.constants import NUM_FBANKS, NUM_FRAMES, CHECKPOINTS_TRIPLET_DIR, BATCH_SIZE
+from deep_speaker.conv_models import DeepSpeakerModel
+from deep_speaker.eval_metrics import evaluate
+from deep_speaker.utils import load_best_checkpoint, enable_deterministic
 
 logger = logging.getLogger(__name__)
 

train.py → deep_speaker/train.py

@@ -5,11 +5,11 @@
 from tensorflow.keras.optimizers import SGD
 from tqdm import tqdm
 
-from batcher import KerasFormatConverter, LazyTripletBatcher
-from constants import BATCH_SIZE, CHECKPOINTS_SOFTMAX_DIR, CHECKPOINTS_TRIPLET_DIR, NUM_FRAMES, NUM_FBANKS
-from conv_models import DeepSpeakerModel
-from triplet_loss import deep_speaker_loss
-from utils import load_best_checkpoint, ensures_dir
+from deep_speaker.batcher import KerasFormatConverter, LazyTripletBatcher
+from deep_speaker.constants import BATCH_SIZE, CHECKPOINTS_SOFTMAX_DIR, CHECKPOINTS_TRIPLET_DIR, NUM_FRAMES, NUM_FBANKS
+from deep_speaker.conv_models import DeepSpeakerModel
+from deep_speaker.triplet_loss import deep_speaker_loss
+from deep_speaker.utils import load_best_checkpoint, ensures_dir
 
 logger = logging.getLogger(__name__)
 

utils.py → deep_speaker/utils.py

@@ -10,7 +10,7 @@
 import pandas as pd
 from natsort import natsorted
 
-from constants import TRAIN_TEST_RATIO
+from deep_speaker.constants import TRAIN_TEST_RATIO
 
 logger = logging.getLogger(__name__)
 

example.py

@@ -2,11 +2,11 @@
 
 import numpy as np
 
-from audio import read_mfcc
-from batcher import sample_from_mfcc
-from constants import SAMPLE_RATE, NUM_FRAMES
-from conv_models import DeepSpeakerModel
-from test import batch_cosine_similarity
+from deep_speaker.audio import read_mfcc
+from deep_speaker.batcher import sample_from_mfcc
+from deep_speaker.constants import SAMPLE_RATE, NUM_FRAMES
+from deep_speaker.conv_models import DeepSpeakerModel
+from deep_speaker.test import batch_cosine_similarity
 
 # Reproducible results.
 np.random.seed(123)

example_pcm.py

@@ -2,9 +2,9 @@
 import numpy as np
 import tensorflow as tf
 
-from constants import SAMPLE_RATE
-from conv_models import DeepSpeakerModel
-from test import batch_cosine_similarity
+from deep_speaker.constants import SAMPLE_RATE
+from deep_speaker.conv_models import DeepSpeakerModel
+from deep_speaker.test import batch_cosine_similarity
 
 # Define the model here.
 model = DeepSpeakerModel(pcm_input=True)

setup.py

@@ -0,0 +1,39 @@
+import os
+import platform
+
+from setuptools import setup
+
+tensorflow = 'tensorflow'
+if platform.system() == 'Darwin' and platform.processor() == 'arm':
+    tensorflow = 'tensorflow-macos'
+    # https://github.com/grpc/grpc/issues/25082
+    os.environ['GRPC_PYTHON_BUILD_SYSTEM_OPENSSL'] = '1'
+    os.environ['GRPC_PYTHON_BUILD_SYSTEM_ZLIB'] = '1'
+
+# numba>=0.48
+install_requires = [
+    tensorflow,
+    'numpy',
+    'natsort',
+    'librosa',
+    'dill',
+    'python_speech_features',
+    'tqdm',
+    'click',
+    'pandas',
+    'matplotlib',
+    'keras',
+    'numba'
+]
+
+setup(
+    name='deep-speaker',
+    version='1.0',
+    description='Deep Speaker',
+    author='Philippe Remy',
+    license='MIT',
+    long_description_content_type='text/markdown',
+    long_description=open('README.md').read(),
+    packages=['deep_speaker'],
+    install_requires=install_requires
+)