def create(entity, entity_name, categories, mutability, **kwargs):
"""This command will create the workspace structure with data and spec file for an entity and set the store configurations.
Example:
create('dataset', 'dataset-ex', categories=['computer-vision', 'images'], mutability='strict')
Args:
entity (str): The type of an ML entity. (dataset, labels or model).
entity_name (str): An ml-git entity name to identify a ML entity.
categories (list): Artifact's category name.
mutability (str): Mutability type. The mutability options are strict, flexible and mutable.
store_type (str, optional): Data store type [default: s3h].
version (int, optional): Number of retries to upload the files to the storage [default: 2].
import_path (str, optional): Path to be imported to the project.
bucket_name (str, optional): Bucket name.
import_url (str, optional): Import data from a google drive url.
credentials_path (str, optional): Directory of credentials.json.
unzip (bool, optional): Unzip imported zipped files [default: False].
"""
args = {'artifact_name': entity_name, 'category': categories, 'mutability': mutability,
'version_number': kwargs.get('version', 1), 'import': kwargs.get('import_path', None),
'store_type': kwargs.get('store_type', StoreType.S3H.value),
'bucket_name': kwargs.get('bucket_name', None), 'unzip': kwargs.get('unzip', False),
'import_url': kwargs.get('import_url', None), 'credentials_path': kwargs.get('credentials_path', None),
'wizard_config': False}
repo = Repository(config_load(), entity)
repo.create(args)
def remote_add(entity, remote_url, global_configuration=False):
"""This command will add a remote to store the metadata from this ml-git project.
Examples:
remote_add('dataset', 'https://[email protected]/mlgit-datasets')
Args:
entity (str): The type of an ML entity. (repository, dataset, labels or model).
remote_url(str): URL of an existing remote git repository.
global_configuration (bool, optional): Use this option to set configuration at global level [default: False].
"""
repo = Repository(config_load(), entity)
repo.repo_remote_add(entity, remote_url, global_configuration)
def push(entity, entity_name, retries=2, clear_on_fail=False):
"""This command allows pushing the data of a specific version of an ML entity.
Example:
push('dataset', 'dataset-ex')
Args:
entity (str): The type of an ML entity. (dataset, labels or model).
entity_name (str): An ml-git entity name to identify a ML entity.
retries (int, optional): Number of retries to upload the files to the storage [default: 2].
clear_on_fail (bool, optional): Remove the files from the store in case of failure during the push operation [default: False].
"""
repo = Repository(config_load(), entity)
repo.push(entity_name, retries, clear_on_fail)
def add(entity_type, entity_name, bumpversion=False, fsck=False, file_path=[]):
"""This command will add all the files under the directory into the ml-git index/staging area.
Example:
add('dataset', 'dataset-ex', bumpversion=True)
Args:
entity_type (str): The type of an ML entity. (dataset, labels or model)
entity_name (str): The name of the ML entity you want to add the files.
bumpversion (bool, optional): Increment the entity version number when adding more files [default: False].
fsck (bool, optional): Run fsck after command execution [default: False].
file_path (list, optional): List of files that must be added by the command [default: all files].
"""
repo = Repository(config_load(), entity_type)
repo.add(entity_name, file_path, bumpversion, fsck)
def get_repository_instance(repo_type):
project_repo_type = 'project'
if repo_type not in EntityType.to_list(
) and repo_type != project_repo_type:
raise RuntimeError(output_messages['ERROR_INVALID_ENTITY_TYPE'] %
EntityType.to_list())
return Repository(config_load(), repo_type)
def test_tag_exist(self):
mdpath = os.path.join(self.test_dir, 'metadata')
specpath = 'dataset-ex'
ensure_path_exists(os.path.join(mdpath, specpath))
shutil.copy('hdata/dataset-ex.spec',
os.path.join(mdpath, specpath) + '/dataset-ex.spec')
manifestpath = os.path.join(os.path.join(mdpath, specpath),
'MANIFEST.yaml')
yaml_save(files_mock, manifestpath)
config['mlgit_path'] = self.test_dir
m = Metadata(specpath, mdpath, config, repotype)
r = Repository(config, repotype)
r.init()
fullmetadatapath, categories_subpath, metadata = m.tag_exists(
self.test_dir)
self.assertFalse(metadata is None)
def test_get_tag(self):
mdpath = os.path.join(self.test_dir, 'metadata')
specpath = 'dataset-ex'
ensure_path_exists(os.path.join(mdpath, specpath))
shutil.copy('hdata/dataset-ex.spec', os.path.join(mdpath, specpath) + '/dataset-ex.spec')
manifestpath = os.path.join(os.path.join(mdpath, specpath), 'MANIFEST.yaml')
yaml_save(files_mock, manifestpath)
config['mlgit_path'] = self.test_dir
m = Metadata(specpath, mdpath, config, DATASETS)
r = Repository(config, DATASETS)
r.init()
tag_list = ['computer__images__dataset-ex__1']
with mock.patch('ml_git.metadata.Metadata.list_tags', return_value=tag_list):
target_tag = m.get_tag(specpath, -1)
self.assertEqual(target_tag, tag_list[0])
clear(m.path)
def init(entity):
"""This command will start the ml-git entity.
Examples:
init('repository')
init('dataset')
Args:
entity (str): The type of entity that will be initialized (repository, dataset, labels or model).
"""
if entity == 'repository':
init_mlgit()
elif entity in EntityType.to_list():
repo = Repository(config_load(), entity)
repo.init()
else:
log.error('The type of entity entered is invalid. Valid types are: [repository, dataset, labels or model]')
def clone(repository_url, folder=None, track=False):
"""This command will clone minimal configuration files from repository-url with valid .ml-git/config.yaml,
then initialize the metadata according to configurations.
Example:
clone('https://[email protected]/mlgit-repository')
Args:
repository_url (str): The git repository that will be cloned.
folder (str, optional): Directory that can be created to execute the clone command [default: current path].
track (bool, optional): Set if the tracking of the cloned repository should be kept [default: False].
"""
repo = Repository(config_load(), 'project')
if folder is not None:
repo.clone_config(repository_url, folder, track)
else:
current_directory = os.getcwd()
with tempfile.TemporaryDirectory(dir=current_directory) as tempdir:
mlgit_path = os.path.join(tempdir, 'mlgit')
repo.clone_config(repository_url, mlgit_path, track)
if not os.path.exists(os.path.join(current_directory, '.ml-git')):
shutil.move(os.path.join(mlgit_path, '.ml-git'),
current_directory)
os.chdir(current_directory)
def checkout(entity, tag, sampling=None, retries=2, force=False, dataset=False, labels=False, version=-1):
"""This command allows retrieving the data of a specific version of an ML entity.
Example:
checkout('dataset', 'computer-vision__images3__imagenet__1')
Args:
entity (str): The type of an ML entity. (dataset, labels or model)
tag (str): An ml-git tag to identify a specific version of an ML entity.
sampling (dict): group: <amount>:<group> The group sample option consists of amount and group used to
download a sample.\n
range: <start:stop:step> The range sample option consists of start, stop and step used
to download a sample. The start parameter can be equal or greater than zero. The
stop parameter can be 'all', -1 or any integer above zero.\n
random: <amount:frequency> The random sample option consists of amount and frequency
used to download a sample.
seed: The seed is used to initialize the pseudorandom numbers.
retries (int, optional): Number of retries to download the files from the storage [default: 2].
force (bool, optional): Force checkout command to delete untracked/uncommitted files from the local repository [default: False].
dataset (bool, optional): If exist a dataset related with the model or labels, this one must be downloaded [default: False].
labels (bool, optional): If exist labels related with the model, they must be downloaded [default: False].
Returns:
str: Return the path where the data was checked out.
"""
repo = Repository(config_load(), entity)
repo.update()
if sampling is not None and not validate_sample(sampling):
return None
options = {}
options['with_dataset'] = dataset
options['with_labels'] = labels
options['retry'] = retries
options['force'] = force
options['bare'] = False
options['version'] = version
repo.checkout(tag, sampling, options)
data_path = os.path.join(entity, *tag.split('__')[:-1])
if not os.path.exists(data_path):
data_path = None
return data_path
def get_repository_instance(repo_type):
return Repository(config_load(), repo_type)
def init_repository(entity_type=DATASETS):
return Repository(config_load(), entity_type)
def init_repository(entity_type='dataset'):
return Repository(config_load(), entity_type)
def init_repository(entity_type=DATASETS):
return Repository(config_load(hide_logs=True), entity_type)
