def i2v_tag_download():
"""Download a pretrained i2v network.
Returns
-------
TYPE
Description
"""
model = download('https://s3.amazonaws.com/cadl/models/illust2vec_tag.tfmodel')
tags = download('https://s3.amazonaws.com/cadl/models/tag_list.json')
return model, tags
def test_melody_rnn_generation():
primer_midi = download('https://s3.amazonaws.com/cadl/share/21150_Harry-Potter-and-the-Philosophers-Stone.mid')
download('https://s3.amazonaws.com/cadl/models/basic_rnn.mag')
magenta_utils.synthesize(primer_midi)
fname = 'primer.mid'
assert(os.path.exists('primer.mid'))
generated_primer = midi_io.midi_file_to_sequence_proto(fname)
assert(len(generated_primer.notes) == 14)
fname = 'synthesis.mid'
assert(os.path.exists('synthesis.mid'))
generated_synthesis = midi_io.midi_file_to_sequence_proto(fname)
assert(len(generated_synthesis.notes) == 243)
def celeb_vaegan_download():
"""Download a pretrained celeb vae/gan network.
Returns
-------
TYPE
Description
"""
# Load the model and labels
model = download(
'https://s3.amazonaws.com/cadl/models/celeb.vaegan.tfmodel')
labels = download(
'https://s3.amazonaws.com/cadl/celeb-align/list_attr_celeba.txt')
return model, labels
def test_alice(max_iter=5):
"""Summary
Parameters
----------
max_iter : int, optional
Description
Returns
-------
TYPE
Description
"""
utils.download('https://s3.amazonaws.com/cadl/models/alice.txt.gz')
with gzip.open('alice.txt.gz', 'rb') as fp:
txt = fp.read().decode('utf-8')
return train(txt, n_layers=2, n_cells=20, max_iter=max_iter)
def test_alice(max_iter=5):
"""Summary
Parameters
----------
max_iter : int, optional
Description
Returns
-------
TYPE
Description
"""
utils.download('https://s3.amazonaws.com/cadl/models/alice.txt.gz')
with gzip.open('alice.txt.gz', 'rb') as fp:
txt = fp.read().decode('utf-8')
return train(txt, n_layers=2, n_cells=20, max_iter=max_iter)
def i2v_download():
"""Download a pretrained i2v network.
Returns
-------
TYPE
Description
"""
model = download('https://s3.amazonaws.com/cadl/models/illust2vec.tfmodel')
return model
def test_generation():
download('https://s3.amazonaws.com/cadl/models/model.ckpt-200000.data-00000-of-00001')
download('https://s3.amazonaws.com/cadl/models/model.ckpt-200000.index')
download('https://s3.amazonaws.com/cadl/models/model.ckpt-200000.meta')
wav_file = download('https://s3.amazonaws.com/cadl/share/trumpet.wav')
res = nsynth.synthesize(wav_file, synth_length=100)
max_idx = np.max(np.where(res))
assert(max_idx > 90 and max_idx < 100)
assert(np.max(res) > 0.0)
def test_generation():
download(
'https://s3.amazonaws.com/cadl/models/model.ckpt-200000.data-00000-of-00001'
)
download('https://s3.amazonaws.com/cadl/models/model.ckpt-200000.index')
download('https://s3.amazonaws.com/cadl/models/model.ckpt-200000.meta')
wav_file = download('https://s3.amazonaws.com/cadl/share/trumpet.wav')
res = nsynth.synthesize(wav_file, synth_length=100)
max_idx = np.max(np.where(res))
assert (max_idx > 90 and max_idx < 100)
assert (np.max(res) > 0.0)
def test_trump(max_iter=100):
"""Summary
Parameters
----------
max_iter : int, optional
Description
"""
utils.download(
'https://s3.amazonaws.com/cadl/models/trump.data-00000-of-00001')
utils.download('https://s3.amazonaws.com/cadl/models/trump.meta')
utils.download('https://s3.amazonaws.com/cadl/models/trump.index')
utils.download('https://s3.amazonaws.com/cadl/models/trump.txt')
with open('trump.txt', 'r') as fp:
txt = fp.read()
#train(txt, ckpt_name='trump', max_iter=max_iter)
print(infer(txt, ckpt_name='./trump', n_iterations=max_iter))
def test_trump(max_iter=100):
"""Summary
Parameters
----------
max_iter : int, optional
Description
"""
utils.download(
'https://s3.amazonaws.com/cadl/models/trump.data-00000-of-00001')
utils.download('https://s3.amazonaws.com/cadl/models/trump.meta')
utils.download('https://s3.amazonaws.com/cadl/models/trump.index')
utils.download('https://s3.amazonaws.com/cadl/models/trump.txt')
with open('trump.txt', 'r') as fp:
txt = fp.read()
#train(txt, ckpt_name='trump', max_iter=max_iter)
print(infer(txt, ckpt_name='./trump', n_iterations=max_iter))
def get_model():
"""Summary
Returns
-------
TYPE
Description
"""
# Download the glove model and open a zip file
file = utils.download('http://nlp.stanford.edu/data/wordvecs/glove.6B.zip')
zf = zipfile.ZipFile(file)
# Collect the words and their vectors
words = []
vectors = []
for l in zf.open("glove.6B.300d.txt"):
t = l.strip().split()
words.append(t[0].decode())
vectors.append(list(map(np.double, t[1:])))
# Store as a lookup table
wordvecs = np.asarray(vectors, dtype=np.double)
word2id = {word: i for i, word in enumerate(words)}
return wordvecs, word2id, words
def get_model():
"""Summary
Returns
-------
TYPE
Description
"""
# Download the glove model and open a zip file
file = utils.download('http://nlp.stanford.edu/data/wordvecs/glove.6B.zip')
zf = zipfile.ZipFile(file)
# Collect the words and their vectors
words = []
vectors = []
for l in zf.open("glove.6B.300d.txt"):
t = l.strip().split()
words.append(t[0].decode())
vectors.append(list(map(np.double, t[1:])))
# Store as a lookup table
wordvecs = np.asarray(vectors, dtype=np.double)
word2id = {word: i for i, word in enumerate(words)}
return wordvecs, word2id, words
def test_convert_to_monophonic():
primer_midi = download('https://s3.amazonaws.com/cadl/share/21150_Harry-Potter-and-the-Philosophers-Stone.mid')
primer_sequence = midi_io.midi_file_to_sequence_proto(primer_midi)
magenta_utils.convert_to_monophonic(primer_sequence)
assert(len(primer_sequence.notes) == 254)
def test_harry_potter():
primer_midi = download('https://s3.amazonaws.com/cadl/share/21150_Harry-Potter-and-the-Philosophers-Stone.mid')
primer_sequence = midi_io.midi_file_to_sequence_proto(primer_midi)
assert(len(primer_sequence.notes) == 282)
assert(len(primer_sequence.time_signatures) == 1)