def main():
parser = argparse.ArgumentParser()
arg = parser.add_argument
arg("--model",
type=str,
default="se_resnext101_dr0.75_512",
help="String model name from models dictionary")
arg("--seed", type=int, default=1234, help="Random seed")
arg("--fold", type=int, default=0, help="Validation fold")
arg("--checkpoints_dir",
type=str,
default="../../checkpoints",
help="Directory for loading model weights")
arg("--from-epoch", type=int, default=1, help="Resume training from epoch")
arg("--to-epochs", type=int, default=15, help="Number of epochs to run")
arg("--debug", type=bool, default=False, help="If the debugging mode")
args = parser.parse_args()
set_seed(args.seed)
weights = f"{args.checkpoints_dir}/{args.model}_fold_{args.fold}/"
predict_test(model_name=args.model,
fold=args.fold,
debug=args.debug,
checkpoints_dir=weights,
save_oof=True,
img_size=IMG_SIZE,
from_epoch=0,
to_epoch=10)
def main():
parser = argparse.ArgumentParser()
arg = parser.add_argument
arg("--model-name",
type=str,
default="resnet101_512",
help="String model name from models dictionary")
arg("--seed", type=int, default=1234, help="Random seed")
arg("--fold", type=int, default=0, help="Validation fold")
args = parser.parse_args()
set_seed(args.seed)
model_names = [
"se_resnext101_dr0.75_512",
"se_resnext101_dr_512",
"se_resnext101_dr_512_without_pretrained",
"se_resnext50_512_dr0.8",
"se_resnext50_512",
"resnet101_512",
"resnet50_512",
]
thresholds = [
0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65,
0.7, 0.75, 0.8, 0.9, 1.0
]
get_metrics(RESULTS_DIR,
model_name=args.model_name,
fold=args.fold,
epochs=19,
thresholds=thresholds,
save_metrics=True)
def main():
parser = argparse.ArgumentParser()
arg = parser.add_argument
arg("--seed", type=int, default=1234, help="Random seed")
arg("--fold", type=int, default=0, help="Validation fold")
arg("--output_dir",
type=str,
default="../../output",
help="Directory for loading results")
arg("--save-pics", type=bool, default=True)
arg("--debug", type=bool, default=False, help="If the debugging mode")
args = parser.parse_args()
set_seed(args.seed)
RESULTS_DIR = "../../../../output/"
model_names = [
"se_resnext101_dr0.75_512",
"se_resnext101_dr_512",
"se_resnext101_512",
"se_resnext50_512_dr0.8",
"se_resnext50_512",
"resnet101_512",
"inc_resnet_v2_512",
]
labels = model_names
model = model_names[0]
all_scores, thresholds, epochs = load_scores(model_name=model,
fold=args.fold,
output_dir=RESULTS_DIR)
scores_heatmap(all_scores, thresholds, epochs, model, RESULTS_DIR)
# plot_best_scores(all_scores, thresholds, epochs,
# model_name=model, output_dir = RESULTS_DIR, if_save=False)
map_2d(all_scores, thresholds, epochs, model, RESULTS_DIR, if_save=False)
plot_score(all_scores,
thresholds,
epochs,
output_dir=RESULTS_DIR,
model_name=model,
if_save=True)
plot_scores(all_scores,
thresholds,
epochs,
model_names,
labels,
RESULTS_DIR,
if_save=True)
def main():
parser = argparse.ArgumentParser()
arg = parser.add_argument
arg("--action",
type=str,
default="train",
help=
"Choose action: train, test_model, check_metric, generate_predictions")
arg("--model",
type=str,
default="se_resnext101_dr0.75_512",
help="String model name from models dictionary")
arg("--run",
type=str,
default="",
help="Experiment id string to be added for saving model")
arg("--seed", type=int, default=1234, help="Random seed")
arg("--fold", type=int, default=0, help="Validation fold")
arg("--weights_dir",
type=str,
default="../../checkpoints",
help="Directory for loading model weights")
arg("--epoch", type=int, default=12, help="Current epoch")
arg("--from-epoch", type=int, default=2, help="Resume training from epoch")
arg("--num-epochs", type=int, default=20, help="Number of epochs to run")
arg("--batch-size", type=int, default=4, help="Batch size for training")
arg("--learning-rate",
type=float,
default=1e-5,
help="Initial learning rate")
arg("--debug", type=bool, default=False, help="If the debugging mode")
args = parser.parse_args()
set_seed(args.seed)
if args.action == "train":
train(
model_name=args.model,
run=args.run,
fold=args.fold,
debug=args.debug,
epochs=args.num_epochs,
)
if args.action == "test_model":
run_str = "" if args.run is None or args.run == "" else f"_{args.run}"
weights = (
f"{WEIGHTS_DIR}/{args.model}{run_str}_fold_{args.fold}/{args.model}_{args.epoch:03}.pt"
)
test_model(
model_name=args.model,
fold=args.fold,
debug=args.debug,
checkpoint=weights,
pics_dir=f"{RESULTS_DIR}/pics",
)
if args.action == "check_metric":
all_scores, thresholds, epochs = check_metric(
model_name=args.model,
run=args.run,
fold=args.fold,
oof_dir=f"{RESULTS_DIR}/oof",
start_epoch=1,
end_epoch=15,
save_metrics=True,
)
if args.action == "generate_predictions":
generate_predictions(
model_name=args.model,
run=args.run,
fold=args.fold,
weights_dir=WEIGHTS_DIR,
debug=args.debug,
from_epoch=args.from_epoch,
to_epoch=args.num_epochs + 1,
save_oof=True,
)