def __init__(self, dataset, model, args):
# Parse key:val arguments
nbcores = len(os.sched_getaffinity(0))
if nbcores == 0:
nbcores = 4 # Arbitrary fallback
args = tools.parse_keyval(args,
defaults={
"batch-size": 32,
"eval-batch-size": 1024,
"weight-decay": 0.00004,
"label-smoothing": 0.,
"labels-offset": 0,
"nb-fetcher-threads": nbcores,
"nb-batcher-threads": nbcores
})
if args["batch-size"] <= 0:
raise tools.UserException(
"Cannot make batches of non-positive size")
# Report experiments
with tools.Context("slim", None):
print("Dataset name in use: " + repr(dataset[0]) + " (in " +
repr(dataset[1]) + ")")
print("Dataset preprocessing: " +
(repr(args["preprocessing"]) if "preprocessing" in
args else "<model default>"))
print("Model name in use: " + repr(model))
# Finalization
self.__args = args
self.__dataset = dataset
self.__preproc = args[
"preprocessing"] if "preprocessing" in args else model
self.__model = model
self.__cntr_wk = 0 # Worker instantiation counter
self.__cntr_ev = 0 # Evaluator instantiation counter
def build(struct, name, select, args, **kwargs):
""" Call the constructor associated with the given selection and the given keyword + parsed arguments.
Args:
struct Structure defining constructors and their respective arguments
name Name of what is built by the constructor
select Constructor to select
args List of "key:value" command line arguments
... Key-value arguments forwarded to the constructor
"""
# Recover constructor and argument structure
if select not in struct:
raise tools.UserException(
"Unknown " + name + " " + repr(select) + ", " +
("no " + name +
" available" if len(struct) == 0 else "expected one of: '" +
("', '").join(struct.keys()) + "'"))
construct, args_struct = struct[select]
# Translate parameters
defaults = {}
for key, val in args_struct.items():
defaults[key] = val[0]
args_parsed = tools.parse_keyval(args, defaults=defaults)
# Instantiate and return
args_kw = {}
for key, val in args_struct.items(
): # Ignore supplementary parameters by using '_struct' instead of '_parsed'
args_kw[args_struct[key][1]] = args_parsed[key]
return construct(**args_kw, **kwargs)
def __init__(self, nbworkers, nbbyzwrks, args):
# Parse key:val arguments
ps = tools.parse_keyval([] if args is None else args,
defaults={"ps": 0.9})["ps"]
if ps <= 0 or ps > 1:
raise tools.UserException("Invalid selection probability, got %s" %
(ps, ))
# Finalization
self._p = ps
self._f = nbbyzwrks
def __init__(self, args):
# Parse key:val arguments
args = tools.parse_keyval(args, defaults={"batch-size": 32})
if args["batch-size"] <= 0:
raise tools.UserException("Cannot make batches of non-positive size")
# Report loading
with tools.Context("mnist", None):
print("Loading MNIST dataset...")
raw_data = tf.keras.datasets.mnist.load_data()
# Finalization
self.__args = args
self.__raw_data = raw_data
self.__datasets = None
self.__cntr_wk = 0 # Worker instantiation counter
self.__cntr_ev = 0 # Evaluator instantiation counter
def __init__(self, args):
# Parse key:val arguments
nbcores = len(os.sched_getaffinity(0))
if nbcores == 0:
nbcores = 4 # Arbitrary fallback
args = tools.parse_keyval(args,
defaults={
"batch-size": 32,
"eval-batch-size": 1024,
"nb-fetcher-threads": nbcores,
"nb-batcher-threads": nbcores
})
if args["batch-size"] <= 0:
raise tools.UserException(
"Cannot make batches of non-positive size")
# Finalization
self.__args = args
self.__preproc = args[
"preprocessing"] if "preprocessing" in args else "cifarnet"
self.__cntr_wk = 0 # Worker instantiation counter
self.__cntr_ev = 0 # Evaluator instantiation counter
default=100,
help="How many training steps between model evaluations, 0 for no evaluation")
parser.add_argument("--user-input-delta",
type=int,
default=0,
help="How many training steps between two prompts for user command inputs, 0 for no user input")
# Parse command line
return parser.parse_args(sys.argv[1:])
with tools.Context("cmdline", "info"):
args = process_commandline()
# Parse additional arguments
for name in ("gar", "attack", "model", "dataset", "loss", "criterion"):
name = f"{name}_args"
keyval = getattr(args, name)
setattr(args, name, dict() if keyval is None else tools.parse_keyval(keyval))
# Count the number of real honest workers
args.nb_honests = args.nb_workers - args.nb_real_byz
if args.nb_honests < 0:
tools.fatal(f"Invalid arguments: there are more real Byzantine workers ({args.nb_real_byz}) than total workers ({args.nb_workers})")
# Check general training parameters
if args.momentum < 0.:
tools.fatal(f"Invalid arguments: negative momentum factor {args.momentum}")
if args.dampening < 0.:
tools.fatal(f"Invalid arguments: negative dampening factor {args.dampening}")
if args.weight_decay < 0.:
tools.fatal(f"Invalid arguments: negative weight decay factor {args.weight_decay}")
# Check the learning rate and associated options
if args.learning_rate <= 0:
tools.fatal(f"Invalid arguments: non-positive learning rate {args.learning_rate}")
if args.learning_rate_decay < 0:
help=
"How many training steps between two prompts for user command inputs, 0 for no user input"
)
# Parse command line
return parser.parse_args(sys.argv[1:])
with tools.Context("cmdline", "info"):
args = process_commandline()
# Parse additional arguments
for name in ("init_multi", "init_mono", "gar", "attack", "model", "loss",
"criterion"):
name = f"{name}_args"
keyval = getattr(args, name)
setattr(args, name,
dict() if keyval is None else tools.parse_keyval(keyval))
# Count the number of real honest workers
args.nb_honests = args.nb_workers - args.nb_real_byz
if args.nb_honests < 0:
tools.fatal(
f"Invalid arguments: there are more real Byzantine workers ({args.nb_real_byz}) than total workers ({args.nb_workers})"
)
# Check the learning rate and associated options
if args.learning_rate_schedule is None:
if args.learning_rate <= 0:
tools.fatal(
f"Invalid arguments: non-positive learning rate {args.learning_rate}"
)
if args.learning_rate_decay < 0:
tools.fatal(
f"Invalid arguments: negative learning rate decay {args.learning_rate_decay}"