def __new__(cls,
name=None,
directory=None,
module_dir=None,
version=None,
**kwargs):
if cls.__name__ == "Module":
if name:
module = cls.init_with_name(
name=name, version=version, **kwargs)
elif directory:
module = cls.init_with_directory(directory=directory, **kwargs)
elif module_dir:
logger.warning(
"Parameter module_dir is deprecated, please use directory to specify the path"
)
if isinstance(module_dir, list) or isinstance(
module_dir, tuple):
directory = module_dir[0]
version = module_dir[1]
else:
directory = module_dir
module = cls.init_with_directory(directory=directory, **kwargs)
CacheUpdater("update_cache", module.name, module.version).start()
else:
if not name and not directory:
directory = os.path.dirname(
os.path.abspath(sys.modules[cls.__module__].__file__))
module = Module.init_with_directory(
directory=directory, **kwargs)
else:
module = object.__new__(cls)
return module
def preinstall_modules(modules):
configs = []
module_exist = {}
if modules is not None:
for module in modules:
module_name = module if "==" not in module else \
module.split("==")[0]
module_version = None if "==" not in module else \
module.split("==")[1]
if module_exist.get(module_name, "") != "":
print(module_name, "==", module_exist.get(module_name),
" will be ignored cause new version is specified.")
configs.pop()
module_exist.update({module_name: module_version})
try:
CacheUpdater("hub_serving_start",
module=module_name,
version=module_version).start()
m = hub.Module(name=module_name, version=module_version)
method_name = m.desc.attr.map.data['default_signature'].s
if method_name == "":
raise RuntimeError("{} cannot be use for "
"predicting".format(module_name))
configs.append({
"module":
module_name,
"version":
m.version,
"category":
str(m.type).split("/")[0].upper()
})
except Exception as err:
print(err, ", start PaddleHub Serving unsuccessfully.")
exit(1)
return configs
def execute(self, argv):
if not argv:
print("ERROR: Please specify a module name.\n")
self.help()
return False
extra = {"command": "install"}
if argv[0].endswith("tar.gz") or argv[0].endswith("phm"):
result, tips, module_dir = default_module_manager.install_module(
module_package=argv[0], extra=extra)
elif os.path.exists(argv[0]) and os.path.isdir(argv[0]):
result, tips, module_dir = default_module_manager.install_module(
module_dir=argv[0], extra=extra)
else:
module_name = argv[0]
module_version = None if "==" not in module_name else module_name.split(
"==")[1]
module_name = module_name if "==" not in module_name else module_name.split(
"==")[0]
CacheUpdater("hub_install", module_name, module_version).start()
result, tips, module_dir = default_module_manager.install_module(
module_name=module_name,
module_version=module_version,
extra=extra)
print(tips)
return True
def execute(self, argv):
if not argv:
print("ERROR: Please specify a module or a model\n")
self.help()
return False
module_name = argv[0]
CacheUpdater("hub_show", module_name).start()
# nlp model
model_info_file = os.path.join(module_name, "info.yml")
if os.path.exists(model_info_file):
self.show_model_info(model_info_file)
return True
cwd = os.getcwd()
module_dir = default_module_manager.search_module(module_name)
module_dir = (os.path.join(cwd, module_name),
None) if not module_dir else module_dir
if not module_dir or not os.path.exists(module_dir[0]):
print("%s is not existed!" % module_name)
return True
self.show_module_info(module_dir)
return True
def download(name,
save_path,
version=None,
decompress=True,
resource_type='Model',
extra={}):
file = os.path.join(save_path, name)
file = os.path.realpath(file)
if os.path.exists(file):
return
if not hub.HubServer()._server_check():
raise ServerConnectionError
search_result = hub.HubServer().get_resource_url(
name, resource_type=resource_type, version=version, extra=extra)
if not search_result:
raise ResourceNotFoundError(name, version)
CacheUpdater("x_download", name, version).start()
url = search_result['url']
with tmp_dir() as _dir:
if not os.path.exists(save_path):
os.makedirs(save_path)
_, _, savefile = default_downloader.download_file(url=url,
save_path=_dir,
print_progress=True)
if tarfile.is_tarfile(savefile) and decompress:
_, _, savefile = default_downloader.uncompress(file=savefile,
print_progress=True)
shutil.move(savefile, file)
def execute(self, argv):
if not argv:
argv = ['.*']
resource_name = argv[0]
CacheUpdater("hub_search", resource_name).start()
extra = {"command": "search"}
resource_list = hub.HubServer().search_resource(resource_name,
resource_type="Module",
extra=extra)
if utils.is_windows():
placeholders = [20, 8, 8, 20]
else:
placeholders = [30, 8, 8, 25]
tp = TablePrinter(
titles=["ResourceName", "Type", "Version", "Summary"],
placeholders=placeholders)
if len(resource_list) == 0:
if hub.HubServer()._server_check() is False:
print(
"Request Hub-Server unsuccessfully, please check your network."
)
for resource_name, resource_type, resource_version, resource_summary in resource_list:
if resource_type == "Module":
colors = ["yellow", None, None, None]
else:
colors = ["light_red", None, None, None]
tp.add_line(contents=[
resource_name, resource_type, resource_version,
resource_summary
],
colors=colors)
print(tp.get_text())
return True
def execute(self, argv):
args = self.parser.parse_args()
if not args.module_name or not args.model_dir:
ConvertCommand.show_help()
return False
self.module = args.module_name
self.version = args.module_version if args.module_version is not None else '1.0.0'
self.src = args.model_dir
if not os.path.isdir(self.src):
print('`{}` is not exists or not a directory path'.format(self.src))
return False
self.dest = args.output_dir if args.output_dir is not None else os.path.join(
'{}_{}'.format(self.module, str(time.time())))
CacheUpdater("hub_convert", self.module, self.version).start()
os.makedirs(self.dest)
with tmp_dir() as _dir:
self._tmp_dir = _dir
self.create_module_py()
self.create_init_py()
self.create_serving_demo_py()
self.create_module_tar()
print('The converted module is stored in `{}`.'.format(self.dest))
return True
def __init__(self,
name=None,
module_dir=None,
signatures=None,
module_info=None,
assets=None,
processor=None,
extra_info=None,
version=None):
self.desc = module_desc_pb2.ModuleDesc()
self.program = None
self.assets = []
self.helper = None
self.signatures = {}
self.default_signature = None
self.module_info = None
self.processor = None
self.extra_info = {} if extra_info is None else extra_info
if not isinstance(self.extra_info, dict):
raise TypeError(
"The extra_info should be an instance of python dict")
# cache data
self.last_call_name = None
self.cache_feed_dict = None
self.cache_fetch_dict = None
self.cache_program = None
fp_lock = open(os.path.join(CONF_HOME, 'config.json'))
lock.flock(fp_lock, lock.LOCK_EX)
if name:
self._init_with_name(name=name, version=version)
lock.flock(fp_lock, lock.LOCK_UN)
elif module_dir:
self._init_with_module_file(module_dir=module_dir[0])
lock.flock(fp_lock, lock.LOCK_UN)
name = module_dir[0].split("/")[-1]
version = module_dir[1]
elif signatures:
if processor:
if not issubclass(processor, BaseProcessor):
raise TypeError(
"Processor shoule be an instance of paddlehub.BaseProcessor"
)
if assets:
self.assets = utils.to_list(assets)
# for asset in assets:
# utils.check_path(assets)
self.processor = processor
self._generate_module_info(module_info)
self._init_with_signature(signatures=signatures)
lock.flock(fp_lock, lock.LOCK_UN)
else:
lock.flock(fp_lock, lock.LOCK_UN)
raise ValueError("Module initialized parameter is empty")
CacheUpdater(name, version).start()
def execute(self, argv):
if not argv:
print("ERROR: Please specify a module\n")
self.help()
return False
module_name = argv[0]
CacheUpdater("hub_uninstall", module_name).start()
result, tips = default_module_manager.uninstall_module(
module_name=module_name)
print(tips)
return True
def execute(self, argv):
CacheUpdater("hub_list").start()
all_modules = default_module_manager.all_modules()
if utils.is_windows():
placeholders = [20, 40]
else:
placeholders = [25, 50]
tp = TablePrinter(
titles=["ModuleName", "Path"], placeholders=placeholders)
for module_name, module_dir in all_modules.items():
tp.add_line(contents=[module_name, module_dir[0]])
print(tp.get_text())
return True
def start_bert_serving(args):
if platform.system() != "Linux":
print("Error. Bert Service only support linux.")
return False
if ServingCommand.is_port_occupied("127.0.0.1", args.port) is True:
print("Port %s is occupied, please change it." % args.port)
return False
from paddle_gpu_serving.run import BertServer
bs = BertServer(with_gpu=args.use_gpu)
bs.with_model(model_name=args.modules[0])
CacheUpdater("hub_bert_service",
module=args.modules[0],
version="0.0.0").start()
bs.run(gpu_index=args.gpu, port=int(args.port))
def execute(self, argv):
CacheUpdater("hub_help").start()
hub_command = BaseCommand.command_dict["hub"]
help_text = "\n"
help_text += "Usage:\n"
help_text += "%s <command> [options]\n" % hub_command.name
help_text += "\n"
help_text += "Commands:\n"
for command_name, command in self.get_all_commands().items():
if not command.show_in_help or not command.description:
continue
help_text += " %-15s\t\t%s\n" % (command.name,
command.description)
print(help_text)
return True
def __new__(cls, name=None, directory=None, module_dir=None, version=None):
if cls.__name__ == "Module":
if name:
module = cls.init_with_name(name=name, version=version)
elif directory:
module = cls.init_with_directory(directory=directory)
elif module_dir:
logger.warning(
"Parameter module_dir is deprecated, please use directory to specify the path"
)
if isinstance(module_dir, list) or isinstance(
module_dir, tuple):
directory = module_dir[0]
version = module_dir[1]
else:
directory = module_dir
module = cls.init_with_directory(directory=directory)
CacheUpdater("update_cache", module.name, module.version).start()
else:
module = object.__new__(cls)
return module
def execute(self, argv):
CacheUpdater("hub_clear").start()
result = True
total_file_size = 0
total_file_count = 0
for rootdir, dirs, files in os.walk(self.cache_dir(), topdown=False):
for filename in files:
filename = os.path.join(rootdir, filename)
try:
file_size = os.path.getsize(filename)
file_count = file_num_in_dir(filename)
os.remove(filename)
total_file_size += file_size
total_file_count += file_count
except Exception as e:
result = False
for dirname in dirs:
dirname = os.path.join(rootdir, dirname)
try:
dir_size = os.path.getsize(dirname)
file_count = file_num_in_dir(dirname)
os.rmdir(dirname)
total_file_size += dir_size
total_file_count += file_count
except Exception as e:
result = False
if total_file_count != 0:
print("Clear %d cached files." % total_file_count)
print("Free disk space %s." %
file_size_in_human_format(total_file_size))
else:
if result:
print("No cache to release.")
else:
print("Clear cache failed!")
return result
def preinstall_modules(self):
for key, value in self.modules_info.items():
init_args = value["init_args"]
CacheUpdater("hub_serving_start",
module=key,
version=init_args.get("version", "0.0.0")).start()
if "directory" not in init_args:
init_args.update({"name": key})
m = hub.Module(**init_args)
method_name = m.serving_func_name
if method_name is None:
raise RuntimeError("{} cannot be use for "
"predicting".format(key))
exit(1)
category = str(m.type).split("/")[0].upper()
self.modules_info[key].update({
"method_name": method_name,
"code_version": m.code_version,
"version": m.version,
"category": category,
"module": m,
"name": m.name
})
def stop_serving(self, port):
filepath = os.path.join(CONF_HOME, "serving_" + str(port) + ".json")
info = self.load_pid_file(filepath, port)
if info is False:
return
pid = info["pid"]
module = info["module"]
start_time = info["start_time"]
if os.path.exists(filepath):
os.remove(filepath)
if not pid_is_exist(pid):
print("PaddleHub Serving has been stopped.")
return
print("PaddleHub Serving will stop.")
CacheUpdater("hub_serving_stop",
module=module,
addition={
"period_time": time.time() - start_time
}).start()
if platform.system() == "Windows":
os.kill(pid, signal.SIGTERM)
else:
os.killpg(pid, signal.SIGTERM)
def execute(self, argv):
CacheUpdater("hub_version").start()
print("hub %s" % version.hub_version)
return True
def execute(self, argv):
if not argv:
print("ERROR: Please specify a module name.\n")
self.help()
return False
module_name = argv[0]
CacheUpdater("hub_run", module_name).start()
self.parser.prog = '%s %s %s' % (ENTRY, self.name, module_name)
self.arg_input_group = self.parser.add_argument_group(
title="Input options", description="Data input to the module")
self.arg_config_group = self.parser.add_argument_group(
title="Config options",
description=
"Run configuration for controlling module behavior, not required")
self.module = self.find_module(module_name)
if not self.module:
return False
# If the module is not executable, give an alarm and exit
if not self.module.is_runable:
print("ERROR! Module %s is not executable." % module_name)
return False
if self.module.code_version == "v2":
results = self.module.run_func(argv[1:])
else:
self.module.check_processor()
self.add_module_config_arg()
self.add_module_input_arg()
if not argv[1:]:
self.help()
return False
self.args = self.parser.parse_args(argv[1:])
config = self.get_config()
data = self.get_data()
try:
self.check_data(data)
except DataFormatError:
self.help()
return False
results = self.module(sign_name=self.module.default_signature,
data=data,
use_gpu=self.args.use_gpu,
batch_size=self.args.batch_size,
**config)
if six.PY2:
try:
results = json.dumps(results,
encoding="utf8",
ensure_ascii=False)
except:
pass
print(results)
return True
def execute(self, argv):
if not argv:
print("ERROR: Please provide the model/module name\n")
self.help()
return False
mod_name = argv[0]
mod_version = None if "==" not in mod_name else mod_name.split("==")[1]
mod_name = mod_name if "==" not in mod_name else mod_name.split(
"==")[0]
CacheUpdater("hub_download", mod_name, mod_version).start()
self.args = self.parser.parse_args(argv[1:])
self.args.type = self.check_type(self.args.type)
extra = {"command": "download"}
if self.args.type in ["Module", "Model"]:
search_result = hub.HubServer().get_resource_url(
mod_name,
resource_type=self.args.type,
version=mod_version,
extra=extra)
else:
search_result = hub.HubServer().get_resource_url(
mod_name,
resource_type="Module",
version=mod_version,
extra=extra)
self.args.type = "Module"
if search_result == {}:
search_result = hub.HubServer().get_resource_url(
mod_name,
resource_type="Model",
version=mod_version,
extra=extra)
self.args.type = "Model"
url = search_result.get('url', None)
except_md5_value = search_result.get('md5', None)
if not url:
if hub.HubServer()._server_check() is False:
tips = "Request Hub-Server unsuccessfully, please check your network."
else:
tips = "PaddleHub can't find model/module named %s" % mod_name
if mod_version:
tips += " with version %s" % mod_version
tips += ". Please use the 'hub search' command to find the correct model/module name."
print(tips)
return True
need_to_download_file = True
file_name = os.path.basename(url)
file = os.path.join(self.args.output_path, file_name)
if os.path.exists(file):
print("File %s already existed\nWait to check the MD5 value" %
file_name)
file_md5_value = utils.md5_of_file(file)
if except_md5_value == file_md5_value:
print("MD5 check pass.")
need_to_download_file = False
else:
print("MD5 check failed!\nDelete invalid file.")
os.remove(file)
if need_to_download_file:
result, tips, file = default_downloader.download_file(
url=url, save_path=self.args.output_path, print_progress=True)
if not result:
print(tips)
return False
if self.args.uncompress:
result, tips, file = default_downloader.uncompress(
file=file,
dirname=self.args.output_path,
delete_file=True,
print_progress=True)
print(tips)
if self.args.type == "Model":
os.rename(file, "./" + mod_name)
return True
def execute(self, argv):
CacheUpdater("hub_autofinetune").start()
if not argv:
print(
"ERROR: Please specify a script to be finetuned in python.\n")
self.help()
return False
self.fintunee_script = argv[0]
self.parser.prog = '%s %s %s' % (ENTRY, self.name,
self.fintunee_script)
self.arg_params_to_be_searched_group = self.parser.add_argument_group(
title="Input options",
description="Hyperparameters to be searched.")
self.arg_config_group = self.parser.add_argument_group(
title="Autofinetune config options",
description=
"Autofintune configuration for controlling autofinetune behavior, not required"
)
self.arg_finetuned_task_group = self.parser.add_argument_group(
title="Finetuned task config options",
description=
"Finetuned task configuration for controlling finetuned task behavior, not required"
)
self.add_params_file_arg()
self.add_autoft_config_arg()
if not argv[1:]:
self.help()
return False
self.args = self.parser.parse_args(argv[1:])
options_str = ""
if self.args.opts is not None:
options_str = self.convert_to_other_options(self.args.opts)
device_ids = self.args.gpu.strip().split(",")
device_ids = [int(device_id) for device_id in device_ids]
if self.args.evaluator.lower() == "fulltrail":
evaluator = FullTrailEvaluator(self.args.param_file,
self.fintunee_script,
options_str=options_str)
elif self.args.evaluator.lower() == "populationbased":
evaluator = PopulationBasedEvaluator(self.args.param_file,
self.fintunee_script,
options_str=options_str)
else:
raise ValueError("The evaluate %s is not defined!" %
self.args.evaluator)
if self.args.tuning_strategy.lower() == "hazero":
autoft = HAZero(evaluator,
cudas=device_ids,
popsize=self.args.popsize,
output_dir=self.args.output_dir)
elif self.args.tuning_strategy.lower() == "pshe2":
autoft = PSHE2(evaluator,
cudas=device_ids,
popsize=self.args.popsize,
output_dir=self.args.output_dir)
else:
raise ValueError("The tuning strategy %s is not defined!" %
self.args.tuning_strategy)
run_round_cnt = 0
solutions_modeldirs = {}
print("PaddleHub Autofinetune starts.")
while (not autoft.is_stop()) and run_round_cnt < self.args.round:
print("PaddleHub Autofinetune starts round at %s." % run_round_cnt)
output_dir = autoft._output_dir + "/round" + str(run_round_cnt)
res = autoft.step(output_dir)
solutions_modeldirs.update(res)
evaluator.new_round()
run_round_cnt = run_round_cnt + 1
print("PaddleHub Autofinetune ends.")
best_hparams_origin = autoft.get_best_hparams()
best_hparams_origin = autoft.mpi.bcast(best_hparams_origin)
with open(autoft._output_dir + "/log_file.txt", "w") as f:
best_hparams = evaluator.convert_params(best_hparams_origin)
print("The final best hyperparameters:")
f.write("The final best hyperparameters:\n")
for index, hparam_name in enumerate(autoft.hparams_name_list):
print("%s=%s" % (hparam_name, best_hparams[index]))
f.write(hparam_name + "\t:\t" + str(best_hparams[index]) +
"\n")
best_hparams_dir, best_hparams_rank = solutions_modeldirs[tuple(
best_hparams_origin)]
print("The final best eval score is %s." %
autoft.get_best_eval_value())
if autoft.mpi.multi_machine:
print("The final best model parameters are saved as " +
autoft._output_dir + "/best_model on rank " +
str(best_hparams_rank) + " .")
else:
print("The final best model parameters are saved as " +
autoft._output_dir + "/best_model .")
f.write("The final best eval score is %s.\n" %
autoft.get_best_eval_value())
best_model_dir = autoft._output_dir + "/best_model"
if autoft.mpi.rank == best_hparams_rank:
shutil.copytree(best_hparams_dir, best_model_dir)
if autoft.mpi.multi_machine:
f.write(
"The final best model parameters are saved as ./best_model on rank " \
+ str(best_hparams_rank) + " .")
f.write("\t".join(autoft.hparams_name_list) +
"\tsaved_params_dir\trank\n")
else:
f.write(
"The final best model parameters are saved as ./best_model ."
)
f.write("\t".join(autoft.hparams_name_list) +
"\tsaved_params_dir\n")
print(
"The related information about hyperparamemters searched are saved as %s/log_file.txt ."
% autoft._output_dir)
for solution, modeldir in solutions_modeldirs.items():
param = evaluator.convert_params(solution)
param = [str(p) for p in param]
if autoft.mpi.multi_machine:
f.write("\t".join(param) + "\t" + modeldir[0] + "\t" +
str(modeldir[1]) + "\n")
else:
f.write("\t".join(param) + "\t" + modeldir[0] + "\n")
return True
