def train_model(n_iter: int = 100) -> None:
# You can load other languages if specificied in Dockerfile
nlp = spacy.load("en_core_web_sm")
if "ner" not in nlp.pipe_names:
ner = nlp.create_pipe("ner")
nlp.add_pipe(ner)
else:
ner = nlp.get_pipe("ner")
# add labels
for _, annotations in TRAIN_DATA:
for ent in annotations.get("entities"):
ner.add_label(ent[2])
# get names of other pipes to disable them during training
pipe_exceptions = ["ner", "trf_wordpiecer", "trf_tok2vec"]
other_pipes = [
pipe for pipe in nlp.pipe_names if pipe not in pipe_exceptions
]
# only train NER
with nlp.disable_pipes(*other_pipes):
nlp.begin_training()
for step_n in range(n_iter):
random.shuffle(TRAIN_DATA)
losses = {}
batches = minibatch(TRAIN_DATA, size=compounding(1.0, 4.0, 1.001))
for batch in batches:
texts, annotations = zip(*batch)
nlp.update(
texts,
annotations,
drop=0.5,
losses=losses,
)
print("Losses: ", losses)
tracking.log_metrics(step=step_n, **losses)
nlp.to_disk("custom_spacy_model")
def main():
tracking.init()
for i in range(args.steps):
logger.info('Step %s', i)
# Scalars
loss = get_loss(i)
accuracy = get_accuracy(loss)
# training metrics, but don't commit the step.
tracking.log_metrics(step=i, loss=loss, accuracy=accuracy)
# validation metrics, which could be reported in another part of the code
if i % args.validate_every == 0:
tracking.log_metric(name='val_acc', value=accuracy - 0.05, step=i)
# Dist
tracking.log_histogram('distribution', get_dist(i), 'auto', step=i)
# Text
tracking.log_text('text-ex', text=get_text(i), step=i)
# images
log_images(i)
# HTML
tracking.log_html('html-ex', html=get_html(i), step=i)
# Generate sin wave as audio
tracking.log_audio(data=get_audio(i), name='audio', step=i)
time.sleep(0.25)
plot_scatter(100)
get_sin_plot(100)
plot_mpl_figure(100)
log_bokeh(100)
log_altair(100)
log_curves(100)
log_plotly(100)
log_curves(100)
train_network()
tracking.log_data_ref(content=x_train, name='x_train')
tracking.log_data_ref(content=y_train, name='y_train')
tracking.log_data_ref(content=x_test, name='x_test')
tracking.log_data_ref(content=y_test, name='y_test')
logger.info('Transforming data...')
x_train, y_train, x_test, y_test = transform_data(x_train,
y_train,
x_test,
y_test,
args.maxlen)
logger.info('Training...')
score, accuracy = train(max_features=args.max_features,
maxlen=args.maxlen,
epochs=args.epochs,
embedding_size=args.embedding_size,
pool_size=args.pool_size,
kernel_size=args.kernel_size,
filters=args.filters,
lstm_output_size=args.lstm_output_size,
batch_size=args.batch_size,
optimizer=args.optimizer,
log_learning_rate=args.log_learning_rate)
# Polyaxon
tracking.log_metrics(eval_score=score, eval_accuracy=accuracy)
logger.info('Test score: %s', score)
logger.info('Test accuracy: %s', accuracy)
logger.info('Loading data...')
(x_train, y_train), (x_test,
y_test) = imdb.load_data(num_words=args.max_features,
skip_top=args.skip_top,
seed=args.seed)
logger.info('train sequences %s', len(x_train))
logger.info('test sequences %s', len(x_test))
# Polyaxon
tracking.log_data_ref(content=x_train, name='x_train')
tracking.log_data_ref(content=y_train, name='y_train')
tracking.log_data_ref(content=x_test, name='x_test')
tracking.log_data_ref(content=y_test, name='y_test')
logger.info('Transforming data...')
x_train, y_train, x_test, y_test = transform_data(x_train, y_train, x_test,
y_test, args.maxlen)
logger.info('Training...')
accuracy = train(max_features=args.max_features,
maxlen=args.maxlen,
batch_size=args.batch_size,
num_nodes=args.num_nodes,
optimizer=args.optimizer,
log_learning_rate=args.log_learning_rate,
dropout=args.dropout,
epochs=args.epochs)
# Polyaxon
tracking.log_metrics(accuracy=accuracy)
train_iter = mx.io.NDArrayIter(mnist['train_data'], mnist['train_label'], args.batch_size,
shuffle=True)
val_iter = mx.io.NDArrayIter(mnist['test_data'], mnist['test_label'], args.batch_size)
# Polyaxon
tracking.log_data_ref(content=mnist['train_data'], name='x_train')
tracking.log_data_ref(content=mnist['train_label'], name='y_train')
tracking.log_data_ref(content=mnist['test_data'], name='x_test')
tracking.log_data_ref(content=mnist['test_label'], name='y_test')
context = mx.gpu if os.environ.get('NVIDIA_VISIBLE_DEVICES') else mx.cpu
metrics = model(context=context,
train_iter=train_iter,
val_iter=val_iter,
conv1_kernel=args.conv1_kernel,
conv1_filters=args.conv1_filters,
conv1_activation=args.conv1_activation,
conv2_kernel=args.conv1_kernel,
conv2_filters=args.conv1_filters,
conv2_activation=args.conv1_activation,
fc1_hidden=args.fc1_hidden,
fc1_activation=args.fc1_activation,
optimizer=args.optimizer,
log_learning_rate=args.log_learning_rate,
batch_size=args.batch_size,
epochs=args.epochs)
# Polyaxon
tracking.log_metrics(accuracy=metrics)
type=int,
default=2000,
help='The maximum number of features.')
parser.add_argument('--max_df',
type=float,
default=1.0,
help='the maximum document frequency.')
parser.add_argument(
'--C',
type=float,
default=1.0,
help='Inverse of regularization strength of LogisticRegression')
args = parser.parse_args()
# Polyaxon
tracking.init()
# Train and eval the model with given parameters.
# Polyaxon
output_path = os.path.join(tracking.get_outputs_path(), "model.joblib")
metrics = train_and_eval(output=output_path,
ngram_range=(args.ngram, args.ngram),
max_features=args.max_features,
max_df=args.max_df,
C=args.C)
# Logging metrics
print("Testing metrics: {}", metrics)
# Polyaxon
tracking.log_metrics(**metrics)
# Polyaxon
tracking.init()
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()
# Polyaxon
tracking.log_data_ref(content=x_train, name='x_train')
tracking.log_data_ref(content=y_train, name='y_train')
tracking.log_data_ref(content=x_test, name='x_test')
tracking.log_data_ref(content=y_test, name='y_test')
x_train, y_train, x_test, y_test = transform_data(x_train, y_train, x_test,
y_test)
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(256, activation=tf.keras.activations.relu),
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(10, activation=tf.keras.activations.softmax)
])
accuracy = train(conv1_size=args.conv1_size,
conv2_size=args.conv2_size,
dropout=args.dropout,
hidden1_size=args.hidden1_size,
optimizer=args.optimizer,
log_learning_rate=args.log_learning_rate,
epochs=args.epochs)
# Polyaxon
tracking.log_metrics(eval_accuracy=accuracy)