def parse_args():
parser = argparse.ArgumentParser(
description="Train classification models on ImageNet",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
models.add_model_args(parser)
fit.add_fit_args(parser)
data.add_data_args(parser)
dali.add_dali_args(parser)
data.add_data_aug_args(parser)
return parser.parse_args()
def get_args():
parser = ArgumentParser()
base_group = parser.add_argument_group('base')
base_group.add_argument('--nmr',
action='store_true',
help='determine whether use nmr data or not')
base_group.add_argument(
'--drop_smile',
action='store_true',
help='If true, do not use SMILES data for prediction')
add_data_util_args(parser)
add_model_args(parser)
add_progress_args(parser)
add_training_args(parser)
add_learning_rate_option(parser)
return parser.parse_args()
def get_training_args():
parser = argparse.ArgumentParser()
# Dynet
dynn.command_line.add_dynet_args(parser)
# Data
data_group = parser.add_argument_group("Data arguments")
data_group.add_argument("--dataset",
default="sst",
choices=["sst", "amazon"])
data_group.add_argument("--data-dir", default="data")
data_group.add_argument("--reprocess-data", action="store_true")
data_group.add_argument("--lowercase", action="store_true")
# Optimization
optim_group = parser.add_argument_group("Optimization arguments")
optim_group.add_argument("--batch-size", default=150, type=int)
optim_group.add_argument("--max-tokens-per-batch", default=4000, type=int)
optim_group.add_argument("--n-epochs", default=10, type=int)
optim_group.add_argument("--patience", default=2, type=int)
optim_group.add_argument("--lr", default=0.001, type=float)
optim_group.add_argument("--lr-decay", default=0.1, type=float)
# Model
model_group = parser.add_argument_group("Model arguments")
model_group.add_argument("--model-type",
default="sopa",
choices=models.supported_model_types)
model_group.add_argument("--model-file", default="sopa_sst.npz")
model_group.add_argument("--pretrained-embeds", default=None, type=str)
model_group.add_argument("--freeze-embeds", action="store_true")
model_group.add_argument("--normalize-embeds", action="store_true")
# Misc
misc_group = parser.add_argument_group("Miscellaneous arguments")
model_group.add_argument("--verbose", action="store_true")
misc_group.add_argument("--log-file", default=None, type=str)
misc_group.add_argument("--n-explain", default=10, type=int)
misc_group.add_argument("--n-top-contrib", default=10, type=int)
# Parse args to get model type
args, _ = parser.parse_known_args()
# Add model specific arguments
models.add_model_args(args.model_type, parser)
# Parse again to get all arguments
args = parser.parse_args()
return args
def run_training_pipeline(parser):
parser.add_argument("-f",
"--config",
default=False,
type=str,
help="Path to a YAML config file.")
parser.add_argument(
"--optimizer",
default=False,
type=str,
help="Optimizer to be used during training.",
)
parser.add_argument(
"--scheduler",
default=False,
type=str,
help="LR scheduler to be used during training.",
)
parser.add_argument(
"--model",
default=False,
type=str,
help="The estimator architecture we we wish to use.",
)
args, _ = parser.parse_known_args()
if not args.optimizer and not args.scheduler and not args.model:
optimizer, scheduler, model = get_main_args_from_yaml(args)
else:
optimizer = args.optimizer
scheduler = args.scheduler
model = args.model
parser = add_optimizer_args(parser, optimizer)
parser = add_scheduler_args(parser, scheduler)
parser = add_model_args(parser, model)
parser = add_trainer_specific_args(parser)
hparams = load_yaml_args(parser=parser, log=log)
set_seed(hparams.seed)
model = build_model(hparams)
trainer = setup_training(hparams)
if hparams.load_weights:
model.load_weights(hparams.load_weights)
log.info(f"{model.__class__.__name__} train starting:")
trainer.fit(model)
parser.add_argument('--patience_factor', default=2, type=int)
parser.add_argument('--max_patience', default=64, type=int)
parser.add_argument('--min_lr', default=1e-6, type=float)
parser.add_argument('--threshold', default=1e-4, type=float)
# optim.lr_scheduler.common
parser.add_argument('--gamma', default=0.25, type=float)
parser.add_argument('--early_stopping', default=128, type=int)
parser.add_argument(
'--model_summary',
default=False,
action='store_true',
)
# development
parser.add_argument('--devrun', default=False, action='store_true')
parser.add_argument('--nosave', default=False, action='store_true')
tmp_args, _ = parser.parse_known_args()
parser = add_model_args(parser, tmp_args)
args = parser.parse_args()
main(args)