Merge pull request #53 from JohnVinyard/python3

Python3

.travis.yml

@@ -1,7 +1,7 @@
 language: python
 python:
   # We don't actually use the Travis Python, but this keeps it organized.
-  - "2.7"
+  - "3.6"
 before_install:
   - chmod +x setup.sh
   - sudo ./setup.sh
@@ -21,7 +21,7 @@ install:
   - conda update -q conda
   # Useful for debugging any issues with conda
   - conda info -a
-  - conda create -q -n test-environment -c pytorch -c hcc -c conda-forge python=$TRAVIS_PYTHON_VERSION numpy=1.15.3 scipy=1.1.0 cython pytorch=0.4 torchvision libsndfile=1.0.28 libsamplerate=0.1.8 libflac=1.3.1 libogg=1.3.2
+  - conda create -q -n test-environment -c pytorch -c hcc -c conda-forge python=$TRAVIS_PYTHON_VERSION numpy=1.15.3 scipy=1.2.1 cython pytorch=0.4 torchvision libsndfile=1.0.28 libsamplerate=0.1.8 libflac=1.3.1 libogg=1.3.2
   - sudo ldconfig /opt/conda/lib
   - source activate test-environment
   - python setup.py install

Dockerfile

@@ -45,13 +45,13 @@ RUN wget http://www.mega-nerd.com/tmp/libsndfile-1.0.26pre5.tar.gz \
     && cd ..
 
 RUN echo 'export PATH=/opt/conda/bin:$PATH' > /etc/profile.d/conda.sh \
-    && wget https://repo.continuum.io/miniconda/Miniconda-latest-Linux-x86_64.sh \
-    && /bin/bash /Miniconda-latest-Linux-x86_64.sh -b -p /opt/conda \
-    && rm /Miniconda-latest-Linux-x86_64.sh
+    && wget https://repo.continuum.io/miniconda/Miniconda3-4.3.21-Linux-x86_64.sh \
+    && /bin/bash /Miniconda3-4.3.21-Linux-x86_64.sh -b -p /opt/conda \
+    && rm /Miniconda3-4.3.21-Linux-x86_64.sh
 
 ENV PATH /opt/conda/bin:$PATH
 
-RUN conda install -y -c pytorch numpy=1.15.3 scipy=1.1.0 pytorch=0.4
+RUN conda install -y -c pytorch numpy=1.15.3 scipy=1.2.1 pytorch=0.4
 
 RUN pip install zounds
 

docs/source/conf.py

@@ -56,9 +56,9 @@
 master_doc = 'index'
 
 # General information about the project.
-project = u'zounds'
-copyright = u'2017, John Vinyard'
-author = u'John Vinyard'
+project = 'zounds'
+copyright = '2017, John Vinyard'
+author = 'John Vinyard'
 
 # The version info for the project you're documenting, acts as replacement for
 # |version| and |release|, also used in various other places throughout the
@@ -232,8 +232,8 @@
 # (source start file, target name, title,
 #  author, documentclass [howto, manual, or own class]).
 latex_documents = [
-    (master_doc, 'zounds.tex', u'zounds Documentation',
-     u'John Vinyard', 'manual'),
+    (master_doc, 'zounds.tex', 'zounds Documentation',
+     'John Vinyard', 'manual'),
 ]
 
 # The name of an image file (relative to this directory) to place at the top of
@@ -262,7 +262,7 @@
 # One entry per manual page. List of tuples
 # (source start file, name, description, authors, manual section).
 man_pages = [
-    (master_doc, 'zounds', u'zounds Documentation',
+    (master_doc, 'zounds', 'zounds Documentation',
      [author], 1)
 ]
 
@@ -276,7 +276,7 @@
 # (source start file, target name, title, author,
 #  dir menu entry, description, category)
 texinfo_documents = [
-    (master_doc, 'zounds', u'zounds Documentation',
+    (master_doc, 'zounds', 'zounds Documentation',
      author, 'zounds', 'One line description of project.',
      'Miscellaneous'),
 ]

examples/effects.py

@@ -44,11 +44,11 @@ class Sound(BaseModel):
     def samples(x):
         return zounds.AudioSamples(x, sr)
 
-    batch_slow = map(samples, time_stretch(windowed, 0.75))
-    batch_fast = map(samples, time_stretch(windowed, 1.25))
+    batch_slow = list(map(samples, time_stretch(windowed, 0.75)))
+    batch_fast = list(map(samples, time_stretch(windowed, 1.25)))
 
-    batch_higher = map(samples, pitch_shift(windowed, 1.0))
-    batch_lower = map(samples, pitch_shift(windowed, -1.0))
+    batch_higher = list(map(samples, pitch_shift(windowed, 1.0)))
+    batch_lower = list(map(samples, pitch_shift(windowed, -1.0)))
 
     app = zounds.ZoundsApp(
         model=Sound,

examples/erb_mdct.py

@@ -16,7 +16,7 @@
 makes inversion easier, at the cost of more redundancy.
 """
 
-from __future__ import division
+
 import zounds
 import scipy
 

examples/freesound.py

@@ -42,9 +42,9 @@ class Sound(BaseModel):
         url = request.url
         if not Sound.exists(url):
             Sound.process(meta=metadata, _id=url)
-            print 'processed {url}'.format(**locals())
+            print('processed {url}'.format(**locals()))
         else:
-            print 'already processed {url}'.format(**locals())
+            print('already processed {url}'.format(**locals()))
 
     snds = list(Sound)
 

examples/hamming_index.py

@@ -57,7 +57,7 @@ def total_duration(doc, ts):
     if not len(index):
         index.add_all()
 
-    for i in xrange(1000):
+    for i in range(1000):
         list(index.random_search(n_results=50, sort=True))
 
 

examples/mdct_synth.py

@@ -124,21 +124,21 @@ class DctKmeansWithLogAmplitude(ff.BaseModel):
     # https://archive.org/details/FlavioGaete
     # - https://archive.org/download/FlavioGaete/FlavioGaete22.zip
     filename = 'FlavioGaete22.zip'
-    print 'Processing Audio...'
+    print('Processing Audio...')
     for zf in ff.iter_zip(filename):
         if '._' in zf.filename:
             continue
         try:
             # check if the feature already exists
             Document(zf.filename).mdct.shape
         except KeyError:
-            print 'processing {filename}'.format(filename=zf.filename)
+            print('processing {filename}'.format(filename=zf.filename))
             Document.process(meta=zf, _id=zf.filename)
 
-    print 'learn k-means clusters'
+    print('learn k-means clusters')
     DctKmeans.process(docs=(doc.mdct for doc in Document))
 
-    print 'learn k-means clusters with log amplitude'
+    print('learn k-means clusters with log amplitude')
     DctKmeansWithLogAmplitude.process(docs=(doc.mdct for doc in Document))
 
     synth = zounds.MDCTSynthesizer()
@@ -167,7 +167,7 @@ def reconstruction(_id=None):
             doc = Document(_id)
         else:
             doc = choice(docs)
-        print doc._id
+        print(doc._id)
         recon_audio = full_pass(doc.mdct, kmeans.pipeline)
         recon_audio_log_amp = full_pass(doc.mdct, kmeans_log_amplitude.pipeline)
         return doc.ogg[:], recon_audio, recon_audio_log_amp

examples/pytorch_autoencoder_raw_audio.py

@@ -277,7 +277,7 @@ def forward(self, input, target):
     # instantiate the trained pipeline
     pipeline = pipeline_cls()
 
-    snds = filter(lambda snd: args.internet_archive_id in snd._id, Sound)
+    snds = [snd for snd in Sound if args.internet_archive_id in snd._id]
     snd = choice(snds)
     time_slice = zounds.TimeSlice(duration=zounds.Seconds(10))
     encoded = pipeline.pipeline.transform(

examples/pytorch_wgan.py

@@ -1,4 +1,4 @@
-from __future__ import division
+
 import numpy as np
 import featureflow as ff
 import zounds
@@ -14,6 +14,7 @@
 from random import choice
 import torch.nn.functional as F
 from uuid import uuid4
+from functools import reduce
 
 samplerate = zounds.SR11025()
 BaseModel = zounds.resampled(resample_to=samplerate, store_resampled=True)
@@ -146,11 +147,11 @@ def __init__(self):
 
         self.feature_map_sizes = reduce(
             lambda x, y: x + [int(x[-1] * FACTOR)],
-            xrange(len(bands) - 1),
+            range(len(bands) - 1),
             [FIRST_FEATURE_MAP_SIZE])
 
         for i, band in enumerate(bands):
-            for j in xrange(i + 1):
+            for j in range(i + 1):
                 fms = self.feature_map_sizes[j]
                 first_layer = j == 0
                 in_channels = 1 if first_layer else self.feature_map_sizes[
@@ -194,11 +195,11 @@ def __init__(self):
         self.factor = FACTOR
         self.feature_map_sizes = reduce(
             lambda x, y: x + [int(x[-1] * FACTOR)],
-            xrange(len(bands) - 1),
+            range(len(bands) - 1),
             [FIRST_FEATURE_MAP_SIZE])
 
         for i, band in enumerate(bands):
-            for j in xrange(i + 1):
+            for j in range(i + 1):
                 fms = self.feature_map_sizes[j]
                 first_layer = j == 0
                 out_channels = 1 if first_layer else self.feature_map_sizes[
@@ -309,10 +310,10 @@ def try_network():
     c = network.discriminator
 
     result = g(v, debug=True)
-    print result.size()
+    print(result.size())
 
     labels = c(result, debug=True)
-    print labels.size()
+    print(labels.size())
 
 
 @zounds.simple_settings
@@ -395,7 +396,7 @@ def load_and_play():
         glob.glob('*.npy'),
         cmp=lambda x, y: int(os.stat(x).st_ctime - os.stat(y).st_ctime))
     most_recent = files[-1]
-    print 'loading generated examples from', most_recent
+    print('loading generated examples from', most_recent)
     results = np.load(most_recent)
 
     # synthesized = FrequencyDecomposition.synthesize_block(results)
@@ -410,11 +411,11 @@ def load_and_play():
 
 
 def synthetic():
-    for i in xrange(100):
+    for i in range(100):
         duration = zounds.Seconds(np.random.randint(2, 20))
         root = np.random.randint(50, 400)
         hz = [root]
-        for _ in xrange(0):
+        for _ in range(0):
             hz.append(hz[-1] * 2)
         synth = zounds.SineSynthesizer(samplerate)
         s = synth.synthesize(duration, hz)
@@ -453,7 +454,7 @@ def arg_maker(epoch):
         v = Variable(t).cuda()
         samples = network.generator(v).data.cpu().numpy().squeeze()
         np.save('epoch' + str(epoch), samples)
-        print 'saved samples for epoch', epoch
+        print('saved samples for epoch', epoch)
         return dict()
 
     # out_channels = 128
@@ -500,7 +501,7 @@ def arg_maker(epoch):
     p = Gan()
 
     def walk2(steps):
-        for i in xrange(steps):
+        for i in range(steps):
             yield np.random.normal(0, 1, LATENT_DIM)
 
     def listen():
@@ -529,7 +530,7 @@ def test_frequency_decomposition():
     inp = Variable(inp).cuda()
     x = gen(inp)
 
-    print x.size()
+    print(x.size())
 
     disc = FDDiscriminator().cuda()
 
@@ -540,10 +541,10 @@ def test_frequency_decomposition():
     # inp = Variable(inp).cuda()
 
     x = disc(x)
-    print x.size()
+    print(x.size())
 
-    print gen.layers
-    print disc.layers
+    print(gen.layers)
+    print(disc.layers)
 
 
 def tweak():
@@ -578,8 +579,8 @@ def get_magnitudes():
         m = np.abs(b).mean()
         magnitudes.append(m)
         scaled.append(np.abs(b / m).mean())
-    print magnitudes
-    print scaled
+    print(magnitudes)
+    print(scaled)
 
 
 if __name__ == '__main__':

examples/richard_nixon_identifier.py

@@ -176,7 +176,7 @@ def generate_training_and_test_set():
         snds = list(Sound)
 
         # get all sounds where Nixon is the speaker
-        nixon = filter(lambda snd: 'Nixon' in snd.meta['artist'], snds)
+        nixon = [snd for snd in snds if 'Nixon' in snd.meta['artist']]
 
         # get an equal number of speeches by anyone besides Nixon
         not_nixon = filter(

examples/segmentsearch.py

@@ -104,14 +104,14 @@ class WithCodes(WithOnsets):
 
     # learn K-Means centroids from the drum hits
     if not BarkKmeans.exists():
-        print 'learning K-Means clusters'
+        print('learning K-Means clusters')
         BarkKmeans.process(docs=(wo.bark for wo in WithOnsets))
 
     # bark_kmeans = BarkKmeans()
 
     # force the new pooled feature to be computed
     for doc in WithCodes:
-        print doc.pooled.slicedata.shape
+        print(doc.pooled.slicedata.shape)
 
     results = index.random_search(n_results=5)
 

examples/timbre.py

@@ -1,5 +1,5 @@
 import os
-from urlparse import urlparse
+from urllib.parse import urlparse
 import featureflow as ff
 import requests
 import zounds
@@ -85,7 +85,7 @@ def download_zip_archive():
     if not os.path.exists(filename):
         resp = requests.get(url, stream=True)
 
-        print 'Downloading {url} -> {filename}...'.format(**locals())
+        print('Downloading {url} -> {filename}...'.format(**locals()))
 
         with open(filename, 'wb') as f:
             for chunk in resp.iter_content(chunk_size=1000000):
@@ -99,30 +99,30 @@ def download_zip_archive():
 
     zip_filename = download_zip_archive()
 
-    print 'Processing Audio...'
+    print('Processing Audio...')
     for zf in ff.iter_zip(zip_filename):
 
         if '._' in zf.filename:
             continue
 
         try:
-            print 'processing {zf.filename}'.format(**locals())
+            print('processing {zf.filename}'.format(**locals()))
             WithTimbre.process(
                 _id=zf.filename, meta=zf, raise_if_exists=True)
         except ff.ModelExistsError as e:
-            print e
+            print(e)
 
     # learn K-Means centroids
     try:
-        print 'learning K-Means centroids'
+        print('learning K-Means centroids')
         BfccKmeans.process(
             docs=(wt.bfcc for wt in WithTimbre), raise_if_exists=True)
     except ff.ModelExistsError:
         pass
 
     # force the new features to be computed, so they're pushed into the index
     for wc in WithCodes:
-        print wc.bfcc_kmeans_pooled
+        print(wc.bfcc_kmeans_pooled)
 
     app = zounds.ZoundsSearch(
         model=WithTimbre,

requirements.txt

@@ -7,5 +7,5 @@ tornado==4.5.3
 pysoundfile
 matplotlib==1.5.0
 numpy==1.15.3
-scipy==1.1.0
+scipy==1.2.1
 torch==0.4.0