def __init__(self,
job_id: str,
role: str,
party_id: int,
model_id: str = None,
model_version: str = None,
component_name: str = None,
component_module_name: str = None,
task_id: str = None,
task_version: int = None,
job_parameters: RunParameters = None):
self.job_id = job_id
self.role = role
self.party_id = party_id
self.model_id = model_id
self.party_model_id = model_utils.gen_party_model_id(model_id=model_id,
role=role,
party_id=party_id)
self.model_version = model_version
self.pipelined_model = None
if self.party_model_id and self.model_version:
self.pipelined_model = pipelined_model.PipelinedModel(
model_id=self.party_model_id, model_version=self.model_version)
self.component_name = component_name if component_name else job_utils.job_virtual_component_name(
)
self.module_name = component_module_name if component_module_name else job_utils.job_virtual_component_module_name(
)
self.task_id = task_id
self.task_version = task_version
self.job_parameters = job_parameters
def do_load_model():
request_data = request.json
request_data['servings'] = RuntimeConfig.SERVICE_DB.get_urls('servings')
role = request_data['local']['role']
party_id = request_data['local']['party_id']
model_id = request_data['job_parameters']['model_id']
model_version = request_data['job_parameters']['model_version']
party_model_id = model_utils.gen_party_model_id(model_id, role, party_id)
if get_base_config('enable_model_store', False):
pipeline_model = pipelined_model.PipelinedModel(
party_model_id, model_version)
component_parameters = {
'model_id': party_model_id,
'model_version': model_version,
'store_address': ServiceRegistry.MODEL_STORE_ADDRESS,
}
model_storage = get_model_storage(component_parameters)
if pipeline_model.exists() and not model_storage.exists(
**component_parameters):
stat_logger.info(
f'Uploading {pipeline_model.model_path} to model storage.')
model_storage.store(**component_parameters)
elif not pipeline_model.exists() and model_storage.exists(
**component_parameters):
stat_logger.info(
f'Downloading {pipeline_model.model_path} from model storage.')
model_storage.restore(**component_parameters)
if not model_utils.check_if_deployed(role, party_id, model_id,
model_version):
return get_json_result(
retcode=100,
retmsg=
"Only deployed models could be used to execute process of loading. "
"Please deploy model before loading.")
retcode, retmsg = publish_model.load_model(request_data)
try:
if not retcode:
with DB.connection_context():
model = MLModel.get_or_none(
MLModel.f_role == request_data["local"]["role"],
MLModel.f_party_id == request_data["local"]["party_id"],
MLModel.f_model_id == request_data["job_parameters"]
["model_id"], MLModel.f_model_version ==
request_data["job_parameters"]["model_version"])
if model:
model.f_loaded_times += 1
model.save()
except Exception as modify_err:
stat_logger.exception(modify_err)
operation_record(request_data, "load",
"success" if not retcode else "failed")
return get_json_result(retcode=retcode, retmsg=retmsg)
def deploy_homo_model(request_data):
party_model_id = model_utils.gen_party_model_id(
model_id=request_data["model_id"],
role=request_data["role"],
party_id=request_data["party_id"])
model_version = request_data["model_version"]
component_name = request_data['component_name']
service_id = request_data['service_id']
framework_name = request_data.get('framework_name')
model = pipelined_model.PipelinedModel(model_id=party_model_id,
model_version=model_version)
if not model.exists():
return 100, 'Model {} {} does not exist'.format(
party_model_id, model_version), None
# get the model alias from the dsl saved with the pipeline
pipeline = model.read_pipeline_model()
train_dsl = json_loads(pipeline.train_dsl)
if component_name not in train_dsl.get('components', {}):
return 100, 'Model {} {} does not contain component {}'.\
format(party_model_id, model_version, component_name), None
model_alias_list = train_dsl['components'][component_name].get(
'output', {}).get('model')
if not model_alias_list:
return 100, 'Component {} in Model {} {} does not have output model'. \
format(component_name, party_model_id, model_version), None
# currently there is only one model output
model_alias = model_alias_list[0]
converted_model_dir = os.path.join(model.variables_data_path,
component_name, model_alias,
"converted_model")
if not os.path.isdir(converted_model_dir):
return 100, '''Component {} in Model {} {} isn't converted'''.\
format(component_name, party_model_id, model_version), None
# todo: use subprocess?
convert_tool = model.get_homo_model_convert_tool()
if not framework_name:
module_name = train_dsl['components'][component_name].get('module')
buffer_obj = model.read_component_model(component_name, model_alias)
framework_name = convert_tool.get_default_target_framework(
model_contents=buffer_obj, module_name=module_name)
model_object = convert_tool.load_converted_model(
base_dir=converted_model_dir, framework_name=framework_name)
deployed_service = model_deploy(party_model_id, model_version,
model_object, framework_name, service_id,
request_data['deployment_type'],
request_data['deployment_parameters'])
return (
0,
f"An online serving service is started in the {request_data['deployment_type']} system.",
deployed_service)
def convert_homo_model(request_data):
party_model_id = model_utils.gen_party_model_id(
model_id=request_data["model_id"],
role=request_data["role"],
party_id=request_data["party_id"])
model_version = request_data.get("model_version")
model = pipelined_model.PipelinedModel(model_id=party_model_id,
model_version=model_version)
if not model.exists():
return 100, 'Model {} {} does not exist'.format(
party_model_id, model_version), None
with open(model.define_meta_path, "r", encoding="utf-8") as fr:
define_index = yaml.safe_load(fr)
framework_name = request_data.get("framework_name")
detail = []
# todo: use subprocess?
convert_tool = model.get_homo_model_convert_tool()
for key, value in define_index.get("model_proto", {}).items():
if key == 'pipeline':
continue
for model_alias in value.keys():
buffer_obj = model.read_component_model(key, model_alias)
module_name = define_index.get("component_define",
{}).get(key, {}).get('module_name')
converted_framework, converted_model = convert_tool.model_convert(
model_contents=buffer_obj,
module_name=module_name,
framework_name=framework_name)
if converted_model:
converted_model_dir = os.path.join(model.variables_data_path,
key, model_alias,
"converted_model")
os.makedirs(converted_model_dir, exist_ok=True)
saved_path = convert_tool.save_converted_model(
converted_model, converted_framework, converted_model_dir)
detail.append({
"component_name": key,
"model_alias": model_alias,
"converted_model_path": saved_path
})
if len(detail) > 0:
return (
0,
f"Conversion of homogeneous federated learning component(s) in model "
f"{party_model_id}:{model_version} completed. Use export or h**o/deploy "
f"to download or deploy the converted model.", detail)
else:
return 100, f"No component in model {party_model_id}:{model_version} can be converted.", None
def get_predict_conf():
request_data = request.json
model_dir = os.path.join(get_fate_flow_directory(), 'model_local_cache')
model_fp_list = glob.glob(
model_dir +
f"/guest#*#{request_data['model_id']}/{request_data['model_version']}")
if model_fp_list:
fp = model_fp_list[0]
pipeline_model = pipelined_model.PipelinedModel(
fp.split('/')[-2],
fp.split('/')[-1])
pipeline = pipeline_model.read_pipeline_model()
predict_dsl = json_loads(pipeline.inference_dsl)
train_runtime_conf = json_loads(pipeline.train_runtime_conf)
parser = schedule_utils.get_dsl_parser_by_version(
train_runtime_conf.get('dsl_version', '1'))
predict_conf = parser.generate_predict_conf_template(
predict_dsl, train_runtime_conf, request_data['model_id'],
request_data['model_version'])
else:
predict_conf = ''
if predict_conf:
if request_data.get("filename"):
os.makedirs(TEMP_DIRECTORY, exist_ok=True)
temp_filepath = os.path.join(TEMP_DIRECTORY,
request_data.get("filename"))
with open(temp_filepath, "w") as fout:
fout.write(json_dumps(predict_conf, indent=4))
return send_file(open(temp_filepath, "rb"),
as_attachment=True,
attachment_filename=request_data.get("filename"))
else:
return get_json_result(data=predict_conf)
return error_response(
210,
"No model found, please check if arguments are specified correctly.")
def operate_model(model_operation):
request_config = request.json or request.form.to_dict()
job_id = job_utils.generate_job_id()
if model_operation not in [
ModelOperation.STORE, ModelOperation.RESTORE,
ModelOperation.EXPORT, ModelOperation.IMPORT
]:
raise Exception(
'Can not support this operating now: {}'.format(model_operation))
required_arguments = ["model_id", "model_version", "role", "party_id"]
check_config(request_config, required_arguments=required_arguments)
request_config["model_id"] = gen_party_model_id(
model_id=request_config["model_id"],
role=request_config["role"],
party_id=request_config["party_id"])
if model_operation in [ModelOperation.EXPORT, ModelOperation.IMPORT]:
if model_operation == ModelOperation.IMPORT:
try:
file = request.files.get('file')
file_path = os.path.join(TEMP_DIRECTORY, file.filename)
# if not os.path.exists(file_path):
# raise Exception('The file is obtained from the fate flow client machine, but it does not exist, '
# 'please check the path: {}'.format(file_path))
try:
os.makedirs(os.path.dirname(file_path), exist_ok=True)
file.save(file_path)
except Exception as e:
shutil.rmtree(file_path)
raise e
request_config['file'] = file_path
model = pipelined_model.PipelinedModel(
model_id=request_config["model_id"],
model_version=request_config["model_version"])
model.unpack_model(file_path)
pipeline = model.read_component_model('pipeline',
'pipeline')['Pipeline']
train_runtime_conf = json_loads(pipeline.train_runtime_conf)
permitted_party_id = []
for key, value in train_runtime_conf.get('role', {}).items():
for v in value:
permitted_party_id.extend([v, str(v)])
if request_config["party_id"] not in permitted_party_id:
shutil.rmtree(model.model_path)
raise Exception(
"party id {} is not in model roles, please check if the party id is valid."
)
try:
adapter = JobRuntimeConfigAdapter(train_runtime_conf)
job_parameters = adapter.get_common_parameters().to_dict()
with DB.connection_context():
db_model = MLModel.get_or_none(
MLModel.f_job_id == job_parameters.get(
"model_version"),
MLModel.f_role == request_config["role"])
if not db_model:
model_info = model_utils.gather_model_info_data(model)
model_info['imported'] = 1
model_info['job_id'] = model_info['f_model_version']
model_info['size'] = model.calculate_model_file_size()
model_info['role'] = request_config["model_id"].split(
'#')[0]
model_info['party_id'] = request_config[
"model_id"].split('#')[1]
if model_utils.compare_version(
model_info['f_fate_version'], '1.5.1') == 'lt':
model_info['roles'] = model_info.get(
'f_train_runtime_conf', {}).get('role', {})
model_info['initiator_role'] = model_info.get(
'f_train_runtime_conf',
{}).get('initiator', {}).get('role')
model_info['initiator_party_id'] = model_info.get(
'f_train_runtime_conf',
{}).get('initiator', {}).get('party_id')
model_info[
'work_mode'] = adapter.get_job_work_mode()
model_info['parent'] = False if model_info.get(
'f_inference_dsl') else True
model_utils.save_model_info(model_info)
else:
stat_logger.info(
f'job id: {job_parameters.get("model_version")}, '
f'role: {request_config["role"]} model info already existed in database.'
)
except peewee.IntegrityError as e:
stat_logger.exception(e)
operation_record(request_config, "import", "success")
return get_json_result()
except Exception:
operation_record(request_config, "import", "failed")
raise
else:
try:
model = pipelined_model.PipelinedModel(
model_id=request_config["model_id"],
model_version=request_config["model_version"])
if model.exists():
archive_file_path = model.packaging_model()
operation_record(request_config, "export", "success")
return send_file(archive_file_path,
attachment_filename=os.path.basename(
archive_file_path),
as_attachment=True)
else:
operation_record(request_config, "export", "failed")
res = error_response(
response_code=210,
retmsg="Model {} {} is not exist.".format(
request_config.get("model_id"),
request_config.get("model_version")))
return res
except Exception as e:
operation_record(request_config, "export", "failed")
stat_logger.exception(e)
return error_response(response_code=210, retmsg=str(e))
else:
data = {}
job_dsl, job_runtime_conf = gen_model_operation_job_config(
request_config, model_operation)
submit_result = DAGScheduler.submit(
{
'job_dsl': job_dsl,
'job_runtime_conf': job_runtime_conf
},
job_id=job_id)
data.update(submit_result)
operation_record(data=job_runtime_conf,
oper_type=model_operation,
oper_status='')
return get_json_result(job_id=job_id, data=data)
def migration(config_data: dict):
try:
party_model_id = model_utils.gen_party_model_id(
model_id=config_data["model_id"],
role=config_data["local"]["role"],
party_id=config_data["local"]["party_id"])
model = pipelined_model.PipelinedModel(
model_id=party_model_id,
model_version=config_data["model_version"])
if not model.exists():
raise Exception("Can not found {} {} model local cache".format(
config_data["model_id"], config_data["model_version"]))
with DB.connection_context():
if MLModel.get_or_none(MLModel.f_model_version ==
config_data["unify_model_version"]):
raise Exception(
"Unify model version {} has been occupied in database. "
"Please choose another unify model version and try again.".
format(config_data["unify_model_version"]))
model_data = model.collect_models(in_bytes=True)
if "pipeline.pipeline:Pipeline" not in model_data:
raise Exception("Can not found pipeline file in model.")
migrate_model = pipelined_model.PipelinedModel(
model_id=model_utils.gen_party_model_id(
model_id=model_utils.gen_model_id(config_data["migrate_role"]),
role=config_data["local"]["role"],
party_id=config_data["local"]["migrate_party_id"]),
model_version=config_data["unify_model_version"])
# migrate_model.create_pipelined_model()
shutil.copytree(src=model.model_path, dst=migrate_model.model_path)
pipeline = migrate_model.read_component_model('pipeline',
'pipeline')['Pipeline']
# Utilize Pipeline_model collect model data. And modify related inner information of model
train_runtime_conf = json_loads(pipeline.train_runtime_conf)
train_runtime_conf["role"] = config_data["migrate_role"]
train_runtime_conf["initiator"] = config_data["migrate_initiator"]
adapter = JobRuntimeConfigAdapter(train_runtime_conf)
train_runtime_conf = adapter.update_model_id_version(
model_id=model_utils.gen_model_id(train_runtime_conf["role"]),
model_version=migrate_model.model_version)
# update pipeline.pb file
pipeline.train_runtime_conf = json_dumps(train_runtime_conf, byte=True)
pipeline.model_id = bytes(
adapter.get_common_parameters().to_dict.get("model_id"), "utf-8")
pipeline.model_version = bytes(
adapter.get_common_parameters().to_dict().get("model_version"),
"utf-8")
# save updated pipeline.pb file
migrate_model.save_pipeline(pipeline)
shutil.copyfile(
os.path.join(migrate_model.model_path, "pipeline.pb"),
os.path.join(migrate_model.model_path, "variables", "data",
"pipeline", "pipeline", "Pipeline"))
# modify proto
with open(
os.path.join(migrate_model.model_path, 'define',
'define_meta.yaml'), 'r') as fin:
define_yaml = yaml.safe_load(fin)
for key, value in define_yaml['model_proto'].items():
if key == 'pipeline':
continue
for v in value.keys():
buffer_obj = migrate_model.read_component_model(key, v)
module_name = define_yaml['component_define'].get(
key, {}).get('module_name')
modified_buffer = model_migration(
model_contents=buffer_obj,
module_name=module_name,
old_guest_list=config_data['role']['guest'],
new_guest_list=config_data['migrate_role']['guest'],
old_host_list=config_data['role']['host'],
new_host_list=config_data['migrate_role']['host'],
old_arbiter_list=config_data.get('role',
{}).get('arbiter', None),
new_arbiter_list=config_data.get('migrate_role',
{}).get('arbiter', None))
migrate_model.save_component_model(
component_name=key,
component_module_name=module_name,
model_alias=v,
model_buffers=modified_buffer)
archive_path = migrate_model.packaging_model()
shutil.rmtree(os.path.abspath(migrate_model.model_path))
return (0, f"Migrating model successfully. " \
"The configuration of model has been modified automatically. " \
"New model id is: {}, model version is: {}. " \
"Model files can be found at '{}'.".format(adapter.get_common_parameters()["model_id"],
migrate_model.model_version,
os.path.abspath(archive_path)),
{"model_id": migrate_model.model_id,
"model_version": migrate_model.model_version,
"path": os.path.abspath(archive_path)})
except Exception as e:
return 100, str(e), {}
def import_from_files(config: dict):
model = pipelined_model.PipelinedModel(
model_id=config["model_id"], model_version=config["model_version"])
if config['force']:
model.force = True
model.unpack_model(config["file"])
def download_model(request_data):
model = pipelined_model.PipelinedModel(model_id=request_data.get("namespace"),
model_version=request_data.get("name"))
model_data = model.collect_models(in_bytes=True)
return model_data
def download_model(party_model_id, model_version):
model = pipelined_model.PipelinedModel(party_model_id, model_version)
if not model.exists():
return
return model.collect_models(in_bytes=True)
def operate_model(model_operation):
request_config = request.json or request.form.to_dict()
job_id = job_utils.generate_job_id()
if model_operation not in [
ModelOperation.STORE, ModelOperation.RESTORE,
ModelOperation.EXPORT, ModelOperation.IMPORT
]:
raise Exception(
'Can not support this operating now: {}'.format(model_operation))
required_arguments = ["model_id", "model_version", "role", "party_id"]
check_config(request_config, required_arguments=required_arguments)
request_config["model_id"] = gen_party_model_id(
model_id=request_config["model_id"],
role=request_config["role"],
party_id=request_config["party_id"])
if model_operation in [ModelOperation.EXPORT, ModelOperation.IMPORT]:
if model_operation == ModelOperation.IMPORT:
try:
file = request.files.get('file')
file_path = os.path.join(TEMP_DIRECTORY, file.filename)
# if not os.path.exists(file_path):
# raise Exception('The file is obtained from the fate flow client machine, but it does not exist, '
# 'please check the path: {}'.format(file_path))
try:
os.makedirs(os.path.dirname(file_path), exist_ok=True)
file.save(file_path)
except Exception as e:
shutil.rmtree(file_path)
raise e
request_config['file'] = file_path
model = pipelined_model.PipelinedModel(
model_id=request_config["model_id"],
model_version=request_config["model_version"])
model.unpack_model(file_path)
pipeline = model.read_component_model('pipeline',
'pipeline')['Pipeline']
train_runtime_conf = json_loads(pipeline.train_runtime_conf)
permitted_party_id = []
for key, value in train_runtime_conf.get('role', {}).items():
for v in value:
permitted_party_id.extend([v, str(v)])
if request_config["party_id"] not in permitted_party_id:
shutil.rmtree(model.model_path)
raise Exception(
"party id {} is not in model roles, please check if the party id is valid."
)
try:
with DB.connection_context():
model = MLModel.get_or_none(
MLModel.f_job_id == train_runtime_conf[
"job_parameters"]["model_version"],
MLModel.f_role == request_config["role"])
if not model:
MLModel.create(
f_role=request_config["role"],
f_party_id=request_config["party_id"],
f_roles=train_runtime_conf["role"],
f_job_id=train_runtime_conf["job_parameters"]
["model_version"],
f_model_id=train_runtime_conf["job_parameters"]
["model_id"],
f_model_version=train_runtime_conf[
"job_parameters"]["model_version"],
f_initiator_role=train_runtime_conf[
"initiator"]["role"],
f_initiator_party_id=train_runtime_conf[
"initiator"]["party_id"],
f_runtime_conf=train_runtime_conf,
f_work_mode=train_runtime_conf[
"job_parameters"]["work_mode"],
f_dsl=json_loads(pipeline.train_dsl),
f_imported=1,
f_job_status='complete')
else:
stat_logger.info(
f'job id: {train_runtime_conf["job_parameters"]["model_version"]}, '
f'role: {request_config["role"]} model info already existed in database.'
)
except peewee.IntegrityError as e:
stat_logger.exception(e)
operation_record(request_config, "import", "success")
return get_json_result()
except Exception:
operation_record(request_config, "import", "failed")
raise
else:
try:
model = pipelined_model.PipelinedModel(
model_id=request_config["model_id"],
model_version=request_config["model_version"])
if model.exists():
archive_file_path = model.packaging_model()
operation_record(request_config, "export", "success")
return send_file(archive_file_path,
attachment_filename=os.path.basename(
archive_file_path),
as_attachment=True)
else:
operation_record(request_config, "export", "failed")
res = error_response(
response_code=210,
retmsg="Model {} {} is not exist.".format(
request_config.get("model_id"),
request_config.get("model_version")))
return res
except Exception as e:
operation_record(request_config, "export", "failed")
stat_logger.exception(e)
return error_response(response_code=210, retmsg=str(e))
else:
data = {}
job_dsl, job_runtime_conf = gen_model_operation_job_config(
request_config, model_operation)
job_id, job_dsl_path, job_runtime_conf_path, logs_directory, model_info, board_url = DAGScheduler.submit(
{
'job_dsl': job_dsl,
'job_runtime_conf': job_runtime_conf
},
job_id=job_id)
data.update({
'job_dsl_path': job_dsl_path,
'job_runtime_conf_path': job_runtime_conf_path,
'board_url': board_url,
'logs_directory': logs_directory
})
operation_record(data=job_runtime_conf,
oper_type=model_operation,
oper_status='')
return get_json_result(job_id=job_id, data=data)
