def fit(self,
completions,
workdir=None,
batch_size=32,
n_epochs=10,
**kwargs):
"""This is where training happens: train your model given list of completions, then returns dict with created links and resources"""
img_urls, img_classes = [], []
print('Collecting completions...')
for completion in completions:
if is_skipped(completion):
continue
img_urls.append(completion['data'][self.value])
img_classes.append(get_choice(completion))
print('Creating dataset...')
data = ImageClassifierDataset(img_urls, img_classes)
datagen = ImageDataGenerator(featurewise_center=True,
featurewise_std_normalization=True)
print('Training model...')
self.reset_model()
self.model.train(data.images, data.labels, datagen)
print('Save model...')
model_path = os.path.join(workdir, 'model.ckpt')
self.model.save(workdir)
return {'model_path': model_path, 'classes': data.labels}
def fit(self,
completions,
workdir=None,
batch_size=32,
num_epochs=10,
**kwargs):
image_urls, image_classes = [], []
print('Collecting completions...')
for completion in completions:
if is_skipped(completion):
continue
image_urls.append(completion['data'][self.value])
image_classes.append(get_choice(completion))
print('Creating dataset...')
dataset = ImageClassifierDataset(image_urls, image_classes)
dataloader = DataLoader(dataset, shuffle=True, batch_size=batch_size)
print('Train model...')
self.reset_model()
self.model.train(dataloader, num_epochs=num_epochs)
print('Save model...')
model_path = os.path.join(workdir, 'model.pt')
self.model.save(model_path)
return {'model_path': model_path, 'classes': dataset.classes}
def fit(self, completions, workdir=None, **kwargs):
annotations = []
for completion in completions:
if is_skipped(completion):
continue
image_path = get_image_local_path(completion['data'][self.value])
image_label = get_choice(completion)
annotations.append((image_path, image_label))
# Create dataset
ds = tf.data.Dataset.from_tensor_slices(annotations)
def prepare_item(item):
label = tf.argmax(item[1] == self.labels)
img = tf.io.read_file(item[0])
img = tf.image.decode_jpeg(img, channels=3)
img = tf.image.resize(img, [self.image_height, self.image_width])
return img, label
ds = ds.map(prepare_item, num_parallel_calls=tf.data.AUTOTUNE)
ds = ds.cache().shuffle(buffer_size=1000).batch(
self.batch_size).prefetch(buffer_size=tf.data.AUTOTUNE)
self.model.compile(optimizer=tf.keras.optimizers.Adam(),
loss=tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True),
metrics=['acc'])
self.model.fit(ds, epochs=self.epochs)
model_file = os.path.join(workdir, 'checkpoint')
self.model.save_weights(model_file)
return {'model_file': model_file}