class SavedJson:
"""Dict that saves to disk on modifications"""
def __init__(self, filename):
super().__init__()
self.filename = filename
self.fiti = Fitipy(self.filename)
self.data = self.fiti.read().dict()
def __getitem__(self, item):
return self.data[item]
def __setitem__(self, key, value):
changed = self.data.get(key, ...) != value
self.data[key] = value
if changed:
self.fiti.write().dict(self.data)
def __delitem__(self, key):
del self.data[key]
self.fiti.write().dict(self.data)
def get(self, k):
return self.data.get(k)
def update(self, data=None, **kwargs):
self.data.update(data, **kwargs)
self.fiti.write().dict(self.data)
class FilesystemService(ServicePlugin):
def __init__(self, rt, root=None):
ServicePlugin.__init__(self, rt)
self.root = root or expanduser(rt.paths.user_config)
self.fiti = Fitipy(self.root)
if not self.isdir(''):
self.mkdir('')
def read(self, *path) -> FitiReader:
return self.fiti.read(*path)
def write(self, *path) -> FitiWriter:
return self.fiti.write(*path)
def subdir(self, *path):
return FilesystemService(self.rt, join(self.root, *path))
def open(self, *path, mode='r'):
return open(join(self.root, *path), mode)
def isfile(self, *path):
return isfile(join(self.root, *path))
def isdir(self, *path):
return isdir(join(self.root, *path))
def mkdir(self, *path):
makedirs(self.path(*path), exist_ok=True)
def path(self, *path):
return join(self.root, *path)
def __init__(self, parser=None):
parser = parser or ArgumentParser()
add_to_parser(parser, self.usage, True)
args = TrainData.parse_args(parser)
self.args = args = self.process_args(args) or args
if args.invert_samples and not args.samples_file:
parser.error(
'You must specify --samples-file when using --invert-samples')
if args.samples_file and not isfile(args.samples_file):
parser.error('No such file: ' +
(args.invert_samples or args.samples_file))
if not 0.0 <= args.sensitivity <= 1.0:
parser.error('sensitivity must be between 0.0 and 1.0')
output_folder = os.path.join(args.folder, splitext(args.model)[0])
if not os.path.exists(output_folder):
print('Creating output folder:', output_folder)
os.makedirs(output_folder)
args.model = os.path.join(output_folder, args.model)
inject_params(args.model)
save_params(args.model)
self.train, self.test = self.load_data(self.args)
set_loss_bias(1.0 - args.sensitivity)
params = ModelParams(skip_acc=args.no_validation,
extra_metrics=args.extra_metrics)
self.model = create_model(args.model, params)
self.model.summary()
from keras.callbacks import ModelCheckpoint, TensorBoard
checkpoint = ModelCheckpoint(args.model,
monitor=args.metric_monitor,
save_best_only=args.save_best)
epoch_file = splitext(args.model)[0]
epoch_file = os.path.join(epoch_file + '.epoch')
epoch_fiti = Fitipy(epoch_file)
self.epoch = epoch_fiti.read().read(0, int)
def on_epoch_end(a, b):
self.epoch += 1
epoch_fiti.write().write(self.epoch, str)
self.model_base = splitext(self.args.model)[0]
if args.samples_file:
self.samples, self.hash_to_ind = self.load_sample_data(
args.samples_file, self.train)
else:
self.samples = set()
self.hash_to_ind = {}
self.callbacks = [
checkpoint,
TensorBoard(log_dir=self.model_base + '.logs', ),
LambdaCallback(on_epoch_end=on_epoch_end)
]
class SampledTrainer(Trainer):
def __init__(self):
parser = create_parser(usage)
super().__init__(parser)
if self.args.invert_samples:
parser.error('--invert-samples should be left blank')
self.args.samples_file = (self.args.samples_file
or '{model_base}.samples.json').format(
model_base=self.model_base)
self.samples, self.hash_to_ind = self.load_sample_data(
self.args.samples_file, self.train)
self.metrics_fiti = Fitipy(self.model_base + '.logs',
'sampling-metrics.txt')
def write_sampling_metrics(self, predicted):
correct = float(
sum((predicted > 0.5) == (self.train[1] > 0.5)) /
len(self.train[1]))
print('Successfully calculated: {0:.3%}'.format(correct))
lines = self.metrics_fiti.read().lines()
lines.append('{}\t{}'.format(
len(self.samples) / len(self.train[1]), correct))
self.metrics_fiti.write().lines(lines)
def choose_new_samples(self, predicted):
failed_samples = {
calc_sample_hash(inp, target)
for i, (inp, pred, target) in enumerate(
zip(self.train[0], predicted, self.train[1]))
if (pred > 0.5) != (target > 0.5)
}
remaining_failed_samples = failed_samples - self.samples
print('Remaining failed samples:', len(remaining_failed_samples))
return islice(remaining_failed_samples, self.args.num_sample_chunk)
def run(self):
print('Writing to:', self.args.samples_file)
print('Writing metrics to:', self.metrics_fiti.path)
for _ in range(self.args.cycles):
print('Calculating on whole dataset...')
predicted = self.model.predict(self.train[0])
self.samples.update(self.choose_new_samples(predicted))
Fitipy(self.args.samples_file).write().set(self.samples)
print('Added', self.args.num_sample_chunk, 'samples')
self.write_sampling_metrics(predicted)
self.model.fit(*self.sampled_data,
batch_size=self.args.batch_size,
epochs=self.epoch + self.args.epochs,
callbacks=self.callbacks,
initial_epoch=self.epoch,
validation_data=self.test)
def __init__(self, args):
super().__init__(args)
if args.invert_samples and not args.samples_file:
raise ValueError(
'You must specify --samples-file when using --invert-samples')
if args.samples_file and not isfile(args.samples_file):
raise ValueError('No such file: ' +
(args.invert_samples or args.samples_file))
if not 0.0 <= args.sensitivity <= 1.0:
raise ValueError('sensitivity must be between 0.0 and 1.0')
inject_params(args.model)
save_params(args.model)
params = ModelParams(skip_acc=args.no_validation,
extra_metrics=args.extra_metrics,
loss_bias=1.0 - args.sensitivity,
freeze_till=args.freeze_till)
self.model = create_model(args.model, params)
self.train, self.test = self.load_data(self.args)
from keras.callbacks import ModelCheckpoint, TensorBoard
checkpoint = ModelCheckpoint(args.model,
monitor=args.metric_monitor,
save_best_only=args.save_best)
epoch_fiti = Fitipy(splitext(args.model)[0] + '.epoch')
self.epoch = epoch_fiti.read().read(0, int)
def on_epoch_end(_a, _b):
self.epoch += 1
epoch_fiti.write().write(self.epoch, str)
self.model_base = splitext(self.args.model)[0]
if args.samples_file:
self.samples, self.hash_to_ind = self.load_sample_data(
args.samples_file, self.train)
else:
self.samples = set()
self.hash_to_ind = {}
self.callbacks = [
checkpoint,
TensorBoard(log_dir=self.model_base + '.logs', ),
LambdaCallback(on_epoch_end=on_epoch_end)
]
def __init__(self):
parser = create_parser(usage)
self.args = args = TrainData.parse_args(parser)
self.audio_buffer = np.zeros(pr.buffer_samples, dtype=float)
self.vals_buffer = np.zeros(pr.buffer_samples, dtype=float)
params = ModelParams(skip_acc=args.no_validation,
extra_metrics=args.extra_metrics,
loss_bias=1.0 - args.sensitivity)
self.model = create_model(args.model, params)
self.listener = Listener('',
args.chunk_size,
runner_cls=lambda x: None)
from keras.callbacks import ModelCheckpoint, TensorBoard
checkpoint = ModelCheckpoint(args.model,
monitor=args.metric_monitor,
save_best_only=args.save_best)
epoch_fiti = Fitipy(splitext(args.model)[0] + '.epoch')
self.epoch = epoch_fiti.read().read(0, int)
def on_epoch_end(a, b):
self.epoch += 1
epoch_fiti.write().write(self.epoch, str)
self.model_base = splitext(self.args.model)[0]
self.callbacks = [
checkpoint,
TensorBoard(log_dir=self.model_base + '.logs', ),
LambdaCallback(on_epoch_end=on_epoch_end)
]
self.data = TrainData.from_both(args.tags_file, args.tags_folder,
args.folder)
pos_files, neg_files = self.data.train_files
self.neg_files_it = iter(cycle(neg_files))
self.pos_files_it = iter(cycle(pos_files))
class TrainSampledScript(TrainScript):
usage = Usage('''
Train a model, sampling data points with the highest loss from a larger dataset
:-c --cycles int 200
Number of sampling cycles of size {epoch} to run
:-n --num-sample-chunk int 50
Number of new samples to introduce at a time between training cycles
:-sf --samples-file str -
Json file to write selected samples to.
Default = {model_base}.samples.json
:-is --invert-samples
Unused parameter
...
''') | TrainScript.usage
def __init__(self, args):
super().__init__(args)
if self.args.invert_samples:
raise ValueError('--invert-samples should be left blank')
self.args.samples_file = (self.args.samples_file
or '{model_base}.samples.json').format(
model_base=self.model_base)
self.samples, self.hash_to_ind = self.load_sample_data(
self.args.samples_file, self.train)
self.metrics_fiti = Fitipy(self.model_base + '.logs',
'sampling-metrics.txt')
def write_sampling_metrics(self, predicted):
correct = float(
sum((predicted > 0.5) == (self.train[1] > 0.5)) /
len(self.train[1]))
print('Successfully calculated: {0:.3%}'.format(correct))
lines = self.metrics_fiti.read().lines()
lines.append('{}\t{}'.format(
len(self.samples) / len(self.train[1]), correct))
self.metrics_fiti.write().lines(lines)
def choose_new_samples(self, predicted):
failed_samples = {
calc_sample_hash(inp, target)
for i, (inp, pred, target) in enumerate(
zip(self.train[0], predicted, self.train[1]))
if (pred > 0.5) != (target > 0.5)
}
remaining_failed_samples = failed_samples - self.samples
print('Remaining failed samples:', len(remaining_failed_samples))
return islice(remaining_failed_samples, self.args.num_sample_chunk)
def run(self):
print('Writing to:', self.args.samples_file)
print('Writing metrics to:', self.metrics_fiti.path)
for _ in range(self.args.cycles):
print('Calculating on whole dataset...')
predicted = self.model.predict(self.train[0])
self.samples.update(self.choose_new_samples(predicted))
Fitipy(self.args.samples_file).write().set(self.samples)
print('Added', self.args.num_sample_chunk, 'samples')
self.write_sampling_metrics(predicted)
self.model.fit(*self.sampled_data,
batch_size=self.args.batch_size,
epochs=self.epoch + self.args.epochs,
callbacks=self.callbacks,
initial_epoch=self.epoch,
validation_data=self.test)
def main():
args = create_parser(usage).parse_args()
filenames = glob(join(args.folder, '*.wav'))
shuffle(filenames)
wav_id = -1
stream = None
stop_event = Event()
stop_event.set()
p = PyAudio()
atexit.register(p.terminate)
def play_audio(audio_file):
nonlocal stream
if stream:
stop_event.clear()
stop_event.wait()
stream.stop_stream()
stream.close()
stream = None
audio = load_audio(audio_file)[-pr.buffer_samples:]
audio /= 2 * min(audio.mean() + 4 * audio.std(), abs(audio).max())
stream = p.open(format=paFloat32,
channels=1,
rate=pr.sample_rate,
output=True)
stream.start_stream()
def write_audio():
data = audio.astype('float32').tostring()
chunk_size = 1024
for pos in range(chunk_size, len(data) + chunk_size, chunk_size):
if not stop_event.is_set():
stop_event.set()
return
stream.write(data[pos - chunk_size:pos])
while stop_event.is_set():
sleep(chunk_size / pr.sample_rate)
stop_event.set()
Thread(target=write_audio, daemon=True).start()
tags_file = Fitipy(args.tags_file)
tags = tags_file.read().dict()
def submit():
nonlocal wav_id
if wav_id >= 0:
tags[basename(splitext(filenames[wav_id])[0])] = float(
slider.get())
tags_file.write().dict(tags)
wav_id += 1
play_audio(filenames[wav_id])
submit()
master = Tk()
label = Label(master, text='0')
label.pack()
def on_slider_change(x):
label['text'] = str(int(float(x)))
slider = Scale(master, from_=0, to=100, command=on_slider_change)
slider.pack()
Button(master, text='Submit', command=submit).pack()
Button(master,
text='Replay',
command=lambda: play_audio(filenames[wav_id])).pack()
mainloop()
stream.stop_stream()
stream.close()
def main():
args = create_parser(usage).parse_args()
num_seen_file = Fitipy(args.cache_file + '.num')
topics_cache = args.cache_file + '.topics.json'
if not isfile(topics_cache):
print('Generating topics...')
with open(topics_cache, 'w') as f:
json.dump(get_keywords_uiuc(), f)
with open(topics_cache) as f:
topics = json.load(f)
num_seen = num_seen_file.read().read(0, int)
with open(args.auth_file) as f:
auth = yaml.load(f)
email = auth['username']
password = auth['password']
server = auth.get('pop3_host', 'pop3.' + email.split('@')[-1])
client = StatelessClass(EmailReceiver,
email=email,
password=password,
server=server) # type: EmailReceiver
print('Waiting for emails...')
while True:
num_messages = len(client.get_list())
if num_messages < num_seen:
num_seen = num_messages
num_seen_file.write().write(num_seen)
if num_messages <= num_seen:
time.sleep(1)
continue
for msg_id in range(num_seen + 1, num_messages + 1):
email = client.get_email(msg_id)
print('Found new email from {} titled {}.'.format(
email['From'], email['Subject']))
email_txt = '\n'.join(email['text'])
email_txt = BeautifulSoup(email_txt).text
email_txt = re.sub(r'https?://(?:[-\w.]|(?:%[\da-fA-F]{2}))', '',
email_txt)
freq = calc_freq(email_txt, topics)
tags = relevant_topics(freq)
print('Found the following tags:', ', '.join(tags))
events = Event.find()
matched_events = [
event for event in events if event.get('emailSrc') and
SequenceMatcher(a=event['emailSrc'], b=email_txt).ratio() > 0.9
]
if matched_events:
print('Ignoring, similar to {} other emails'.format(
len(matched_events)))
else:
Event.add({
'name': email['Subject'],
'description': email_txt,
'location': '',
'time': int(time.time()),
'tags': tags,
'emailSrc': email_txt
})
num_seen += 1
num_seen_file.write().write(num_seen, str)