def __init__(self, settings):
super(Trainer, self).__init__()
self.settings = settings
self.phase = settings.cmd
self.batch_size = settings.batch_size
self.data_dir = settings.data_dir
self.list_dir = settings.list_dir
self.checkpoint = settings.resume
self.load_checkpoint = (len(self.checkpoint) > 0)
self.num_epochs = settings.num_epochs
self.lr = float(settings.lr)
self.save = settings.save_on or settings.out_dir
self.from_pause = self.settings.continu
self.path_ctrl = settings.global_path
self.path = self.path_ctrl.get_path
log_dir = '' if settings.log_off else self.path_ctrl.get_dir('log')
self.logger = Logger(scrn=True, log_dir=log_dir, phase=self.phase)
for k, v in sorted(settings.__dict__.items()):
self.logger.show("{}: {}".format(k, v))
self.start_epoch = 0
self._init_max_acc = 0.0
self.model = None
self.criterion = None
def set_gpc_and_logger(args):
gpc = OutPathGetter(root=os.path.join(args.exp_dir, args.tag),
suffix=args.suffix)
log_dir = '' if args.log_off else gpc.get_dir('log')
logger = Logger(scrn=True, log_dir=log_dir, phase=args.cmd)
register('GPC', gpc)
register('LOGGER', logger)
return gpc, logger
def __init__(self,
data_dir,
ckp_path,
save_lr=False,
list_dir='',
out_dir='./',
log_dir=''):
super(Predictor, self).__init__()
self.data_dir = data_dir
self.list_dir = list_dir
self.out_dir = out_dir
self.checkpoint = ckp_path
self.save_lr = save_lr
self.logger = Logger(scrn=True, log_dir=log_dir, phase='test')
self.model = None
def __init__(self,
model=None,
mode='folder',
save_dir=None,
scrn=True,
log_dir=None,
cuda_off=False):
self.save_dir = save_dir
self.output = None
if not cuda_off and torch.cuda.is_available():
self.device = torch.device('cuda')
else:
self.device = torch.device('cpu')
assert model is not None, "The model must be assigned"
self.model = self._model_init(model)
if mode not in Predictor.modes:
raise NotImplementedError
self.logger = Logger(scrn=scrn, log_dir=log_dir, phase='predict')
if mode == 'dataloader':
self._predict = partial(self._predict_dataloader,
dataloader=None,
save_dir=save_dir)
elif mode == 'folder':
# self.suffix = ['.jpg', '.png', '.bmp', '.gif', '.npy'] # 支持的图像格式
self._predict = partial(self._predict_folder, save_dir=save_dir)
elif mode == 'list':
self._predict = partial(self._predict_list, save_dir=save_dir)
elif mode == 'file':
self._predict = partial(self._predict_file, save_dir=save_dir)
elif mode == 'data':
self._predict = partial(self._predict_data, save_dir=save_dir)
else:
raise NotImplementedError
#!/usr/bin/python # -*- coding: UTF-8 -*- """ # @Time : 2019/9/9 9:55 # @Author : peng.wang # @Email : [email protected] # @FileName: model.py # @ProjectName :Facility_Location_FangTai """ from gurobipy import * from utils.misc import Logger import pandas as pd # define the log file log = Logger(log_path='../log').logger # define the facility location problem YEAR_DAY = 365 class FacilityLocation(object): """ this class is consist of attributes for problem construction some utils function for dealing with post-process one key function for building the detail model """ def __init__(self, data, config): """ :param data: class of data provide all data used
import pandas as pd
import numpy as np
from sklearn.base import BaseEstimator
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.ensemble import RandomForestRegressor
from sklearn.gaussian_process import GaussianProcessRegressor
from sklearn.gaussian_process.kernels import DotProduct, WhiteKernel
from sklearn.svm import SVR
from sklearn.linear_model import Lasso, Ridge
from sklearn.model_selection import GridSearchCV
#
from utils.misc import Logger
from utils.util import generate_cutoffs
from utils.misc import save_model_to_file, mean_abs_percentage_error, xgb_mape
log = Logger(log_path=os.path.join(os.path.dirname(os.path.dirname(__file__)),
'log')).logger
warnings.filterwarnings("ignore")
xgb_installed = False
lgt_installed = False
try:
import xgboost as xgb
xgb_installed = True
import lightgbm as lgt
lgt_installed = True
except ImportError:
pass
class TrainModel(BaseEstimator):
RESTORE_FROM = "/data/AutoPheno/green/200527/PatchNet/snapshots-fb/LEAF_UNET_B0064_S010700.pth"
SAVE_PRED_EVERY = 1000
SNAPSHOT_DIR = root_dir + 'PatchNet/snapshots'+postfix
IMGSHOT_DIR = root_dir + 'PatchNet/imgshots'+postfix
WEIGHT_DECAY = 0.0005
NUM_EXAMPLES_PER_EPOCH = 13862
NUM_STEPS_PER_EPOCH = math.ceil(NUM_EXAMPLES_PER_EPOCH / float(BATCH_SIZE))
MAX_ITER = max(NUM_EXAMPLES_PER_EPOCH * MAX_EPOCH + 1,
NUM_STEPS_PER_EPOCH * BATCH_SIZE * MAX_EPOCH + 1)
if not os.path.exists(SNAPSHOT_DIR):
os.makedirs(SNAPSHOT_DIR)
if not os.path.exists(IMGSHOT_DIR):
os.makedirs(IMGSHOT_DIR)
LOG_PATH = SNAPSHOT_DIR + "/B"+format(BATCH_SIZE, "04d")+"E"+format(MAX_EPOCH, "04d")+".log"
sys.stdout = Logger(LOG_PATH, sys.stdout)
print(DATA_LIST_PATH)
print("num of epoch:", MAX_EPOCH)
print("RESTORE_FROM:", RESTORE_FROM)
print(NUM_EXAMPLES_PER_EPOCH)
def get_arguments():
"""Parse all the arguments provided from the CLI.
Returns:
A list of parsed arguments.
"""
parser = argparse.ArgumentParser(description="UNet Network")
parser.add_argument("--set-start", default=False)
parser.add_argument("--start-step", default=0, type=int)
def run(try_num, config):
output_dir = f'./dae-out-{try_num}'
if not os.path.exists(output_dir):
os.mkdir(output_dir)
args = get_args()
train_features = pd.read_csv('../input/lish-moa/train_features.csv')
test_features = pd.read_csv('../input/lish-moa/test_features.csv')
if args.debug:
train_features = train_features.loc[:500]
config.update(dict(n_epochs=3, n_folds=2))
all_features = pd.concat([train_features,
test_features]).reset_index(drop=True)
g_features_columns = [
col for col in all_features.columns if col.startswith('g-')
]
c_features_columns = [
col for col in all_features.columns if col.startswith('c-')
]
feature_columns = g_features_columns + c_features_columns
n_features = len(feature_columns)
kfold = MultilabelStratifiedKFold(n_splits=config.n_folds,
random_state=42,
shuffle=True)
logger = Logger()
for fold_index, (train_idx, valid_idx) in enumerate(
kfold.split(all_features.values, all_features.values)):
print('Fold: ', fold_index + 1, flush=True)
x_train = all_features.loc[train_idx]
x_valid = all_features.loc[valid_idx]
model = new_autoencoder(config.model_kind,
n_features=n_features).to(DEVICE)
criterion = nn.MSELoss()
optimizer = torch.optim.Adam(model.parameters(),
lr=config.learning_rate,
weight_decay=1e-5)
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer,
mode='min',
factor=0.1,
patience=3,
eps=1e-4,
verbose=True)
early_stopping = EarlyStopping(patience=10)
best_score = np.inf
for epoch in range(config.n_epochs):
dataset = DaeDataset(x_train,
feature_columns,
noise_ratio=config.noise_ratio)
dataloader = DataLoader(dataset,
batch_size=config.batch_size,
shuffle=True)
train_loss = loop_train(model, criterion, dataloader, optimizer)
dataset = DaeDataset(x_valid,
feature_columns,
noise_ratio=config.noise_ratio)
dataloader = DataLoader(dataset,
batch_size=config.valid_batch_size,
shuffle=False)
valid_loss, _ = loop_valid(model, criterion, dataloader)
scheduler.step(valid_loss)
logger.update({
'fold': fold_index,
'epoch': epoch + 1,
'train_loss': train_loss,
'val_loss': valid_loss
})
print(
f'epoch {epoch + 1}/{config.n_epochs} - train_loss: {train_loss:.5f} - '
+ f'valid_loss: {valid_loss:.5f}',
flush=True)
if valid_loss < best_score:
best_score = valid_loss
torch.save(model.state_dict(),
f'./{output_dir}/dae_fold_weight_{fold_index}.pt')
if early_stopping.should_stop(valid_loss):
print('Early stopping', flush=True)
break
logger.save(f'./{output_dir}/dae_log.csv')
oof_preds = []
for fold_index in range(config.n_folds):
model = new_autoencoder(config.model_kind,
n_features=n_features).to(DEVICE)
model.load_state_dict(
torch.load(f'./{output_dir}/dae_fold_weight_{fold_index}.pt'))
model.eval()
dataset = DaeDataset(all_features,
feature_columns,
noise_ratio=config.noise_ratio)
dataloader = DataLoader(dataset,
batch_size=config.valid_batch_size,
shuffle=False)
loss, preds = loop_valid(model, nn.MSELoss(), dataloader)
logger.update({'fold': fold_index, 'val_loss': loss})
print('Evaluation fold: {} - valid_loss: {:.5f}'.format(
fold_index, loss),
flush=True)
oof_preds.append(preds)
print('A Whole Evaluation Score: {:.5f}'.format(
mean_squared_error(all_features.loc[:, feature_columns].values,
np.mean(oof_preds, axis=0))),
flush=True)
# for i, preds in enumerate(oof_preds):
# create_pred_feature_df(preds, all_features).to_csv(f'./{output_dir}/dae_features_{i}.csv', index=False)
create_pred_feature_df(np.mean(oof_preds, axis=0), all_features).to_csv(
f'./{output_dir}/dae_features_mean.csv', index=False)
References
----------
[1] Narula, S. & Weistroffer,
H. A flexible method for nonlinear multicriteria decision-making problems Systems,
Man and Cybernetics, IEEE Transactions on, 1989 , 19 , 883-887.
'''
import sys,os
example_path=os.path.dirname(os.path.realpath(__file__))
sys.path.append(os.path.join(example_path,".."))
from utils.misc import Logger
sys.stdout = Logger(os.path.splitext(os.path.basename(__file__))[0])
from utils import tui
from method.NAUTILUS import NAUTILUSv1,ENAUTILUS
from optimization.OptimizationMethod import PointSearch
from problem.Problem import PreGeneratedProblem
if __name__ == '__main__':
# SciPy breaks box constraints
method = ENAUTILUS(PreGeneratedProblem(filename=os.path.join(example_path,"AuxiliaryServices.csv")), PointSearch)
zh=tui.iter_enautilus(method)
ci=method.current_iter
# @Author : peng.wang # @Email : [email protected] # @FileName: util.py # @ProjectName :Facility_Location_FangTai """ import pandas as pd import os import xlwings as xw from utils.misc import Logger from pyecharts import options as opts from pyecharts.charts import Geo from pyecharts.globals import ChartType, SymbolType # define the log file RAW_DATA_PATH = os.path.dirname(os.path.dirname(__file__)) log = Logger(log_path=os.path.join(RAW_DATA_PATH, 'log')).logger class DataHandler(object): """ """ def __init__(self, file, config): """ file:文件名 :param file: :param config """ # 读数路径 self._PATH = os.path.join(os.path.join(RAW_DATA_PATH, 'data'), file) self._config = config
def run(try_num, config):
args = get_args()
print('args', args, flush=True)
print('config:', config.to_dict(), flush=True)
set_seed(config.rand_seed)
pretrained_model = f"tf_efficientnet_b3_ns"
model_dir = f'deepinsight-{try_num}'
if not os.path.exists(model_dir):
os.mkdir(model_dir)
train_features = pd.read_csv(f"../input/lish-moa/train_features.csv")
train_targets = pd.read_csv(f"../input/lish-moa/train_targets_scored.csv")
test_features = pd.read_csv(f"../input/lish-moa/test_features.csv")
if config.dae_path:
dae_features = pd.read_csv(config.dae_path)
if args.debug:
train_features = train_features.iloc[:500]
train_targets = train_targets.iloc[:500]
if config.dae_path:
dae_features = pd.concat([dae_features.iloc[:500], dae_features.iloc[-3982:]]).reset_index(drop=True)
config.update(dict(
kfolds=3,
n_epoch=3
))
train_features = train_features.sort_values(by=["sig_id"], axis=0, inplace=False).reset_index(drop=True)
train_targets = train_targets.sort_values(by=["sig_id"], axis=0, inplace=False).reset_index(drop=True)
cat_features_columns = ["cp_dose", 'cp_time']
num_feature_columns = [c for c in train_features.columns
if c != "sig_id" and c not in cat_features_columns + ['cp_type']]
all_features_columns = cat_features_columns + num_feature_columns
target_columns = [c for c in train_targets.columns if c != "sig_id"]
g_feature_columns = [c for c in num_feature_columns if c.startswith("g-")]
c_feature_columns = [c for c in num_feature_columns if c.startswith("c-")]
if config.dae_path:
if config.dae_strategy == 'replace':
train_features, test_features = assign_dae_features(
train_features, test_features, dae_features, len(num_feature_columns))
else:
train_features, test_features, dae_feature_columns = merge_dae_features(
train_features, test_features, dae_features, len(g_feature_columns), len(c_feature_columns))
all_features_columns += dae_feature_columns
train_targets = train_targets.loc[train_features['cp_type'] == 'trt_cp'].reset_index(drop=True)
train_features = train_features.loc[train_features['cp_type'] == 'trt_cp'].reset_index(drop=True)
if config.normalizer == 'rank':
train_features, test_features = normalize(train_features, test_features, num_feature_columns)
for df in [train_features, test_features]:
df['cp_type'] = df['cp_type'].map({'ctl_vehicle': 0, 'trt_cp': 1})
df['cp_dose'] = df['cp_dose'].map({'D1': 0, 'D2': 1})
df['cp_time'] = df['cp_time'].map({24: 0, 48: 0.5, 72: 1})
if config.variance_target_type == 1:
pickle_path = f'{model_dir}/variance_reduction.pkl'
variance_target_features = num_feature_columns
if config.dae_path and config.dae_strategy != 'replace':
variance_target_features += dae_feature_columns
if not os.path.exists(pickle_path):
vt = variance_reduction_fit(train_features, variance_target_features, config.variance_threshold)
save_pickle(vt, pickle_path)
vt = load_pickle(pickle_path)
train_features = variance_reduction_transform(vt, train_features, variance_target_features)
test_features = variance_reduction_transform(vt, test_features, variance_target_features)
print('(variance_reduction) Number of features after applying:', len(train_features.columns), flush=True)
all_features_columns = list(train_features.columns[1:])
skf = MultilabelStratifiedKFold(n_splits=config.kfolds, shuffle=True, random_state=config.rand_seed)
y_labels = np.sum(train_targets.drop("sig_id", axis=1), axis=0).index.tolist()
logger = Logger()
for fold_index, (train_index, val_index) in enumerate(skf.split(train_features, train_targets[y_labels])):
if args.only_pred:
print('Skip training', flush=True)
break
print(f'Fold: {fold_index}', train_index.shape, val_index.shape, flush=True)
X_train = train_features.loc[train_index, all_features_columns].copy().values
y_train = train_targets.iloc[train_index, 1:].copy().values
X_valid = train_features.loc[val_index, all_features_columns].copy().values
y_valid = train_targets.iloc[val_index, 1:].copy().values
if config.normalizer == 'log':
scaler = LogScaler()
if config.norm_apply_all:
scaler.fit(X_train)
X_train = scaler.transform(X_train)
X_valid = scaler.transform(X_valid)
else:
target_features = [i for i, c in enumerate(all_features_columns) if c in num_feature_columns]
non_target_features = [i for i, c in enumerate(all_features_columns) if c not in num_feature_columns]
scaler.fit(X_train[:, target_features])
X_train_tr = scaler.transform(X_train[:, target_features])
X_valid_tr = scaler.transform(X_valid[:, target_features])
X_train = np.concatenate([X_train[:, non_target_features], X_train_tr], axis=1)
X_valid = np.concatenate([X_valid[:, non_target_features], X_valid_tr], axis=1)
save_pickle(scaler, f'{model_dir}/scaler-{fold_index}.pkl')
transformer = DeepInsightTransformer(
feature_extractor=config.extractor,
pixels=config.resolution,
perplexity=config.perplexity,
random_state=config.rand_seed,
n_jobs=-1
).fit(X_train)
save_pickle(transformer, f'{model_dir}/transformer-{fold_index}.pkl')
model = MoAEfficientNet(
pretrained_model_name=pretrained_model,
fc_size=config.fc_size,
drop_rate=config.drop_rate,
drop_connect_rate=config.drop_connect_rate,
weight_init='goog',
).to(DEVICE)
if config.smoothing is not None:
if config.weighted_loss_weights is not None:
indices = get_minority_target_index(train_targets, threshold=config.weighted_loss_threshold)
indices = [int(i not in indices) for i, c in enumerate(target_columns)]
train_loss_function = SmoothBCEwLogits(
smoothing=config.smoothing,
weight=config.weighted_loss_weights,
weight_targets=indices,
n_labels=len(target_columns))
else:
train_loss_function = SmoothBCEwLogits(smoothing=config.smoothing)
else:
train_loss_function = bce_loss
eval_loss_function = bce_loss
optimizer = optim.Adam(model.parameters(), weight_decay=config.weight_decay, lr=config.learning_rate)
if config.scheduler_type == 'ca':
scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=config.t_max, eta_min=0, last_epoch=-1)
elif config.scheduler_type == 'ms':
scheduler = optim.lr_scheduler.MultiStepLR(optimizer, milestones=config.ms_scheduler_milestones, gamma=0.1)
else:
scheduler = optim.lr_scheduler.ReduceLROnPlateau(
optimizer, mode='min', factor=0.1, patience=config.rp_patience, eps=1e-4, verbose=True)
early_stopping = EarlyStopping(patience=7)
best_score = np.inf
start_time = time.time()
for epoch in range(config.n_epoch):
if config.swap_enable:
dataset = MoAImageSwapDataset(
X_train,
y_train,
transformer,
image_size=config.image_size,
swap_prob=config.swap_prob,
swap_portion=config.swap_portion)
else:
dataset = MoAImageDataset(X_train, y_train, transformer, image_size=config.image_size)
dataloader = DataLoader(
dataset,
batch_size=config.batch_size,
shuffle=True,
num_workers=8,
pin_memory=True,
drop_last=False)
loss = loop_train(model, train_loss_function, dataloader, optimizer)
if config.scheduler_type == 'rp':
scheduler.step(loss)
else:
scheduler.step()
for param_group in optimizer.param_groups:
print('current learning rate:', param_group['lr'])
del dataset, dataloader
dataset = MoAImageDataset(X_valid, y_valid, transformer, image_size=config.image_size)
dataloader = DataLoader(
dataset,
batch_size=config.infer_batch_size,
shuffle=False,
num_workers=8,
pin_memory=True,
drop_last=False)
valid_loss, valid_preds = loop_valid(model, eval_loss_function, dataloader)
del dataset, dataloader
logger.update({'fold': fold_index, 'epoch': epoch + 1, 'train_loss': loss, 'val_loss': valid_loss})
print(f'epoch {epoch + 1}/{config.n_epoch} - train_loss: {loss:.5f} - ' +
f'valid_loss: {valid_loss:.5f} - elapsed: {time_format(time.time() - start_time)}', flush=True)
if valid_loss < best_score:
best_score = valid_loss
torch.save(model.state_dict(), f'./{model_dir}/deepinsight-{fold_index}.pt')
if early_stopping.should_stop(valid_loss):
print('Early stopping', flush=True)
break
print(f'Done -> Fold {fold_index}/{config.kfolds} - best_valid_loss: {best_score:.5f} - ' +
f'elapsed: {time_format(time.time() - start_time)}', flush=True)
torch.cuda.empty_cache()
gc.collect()
if args.return_first_fold:
logger.save(f'{model_dir}/log.csv')
return
test_preds = np.zeros((test_features.shape[0], len(target_columns)))
start_time = time.time()
print('Start infarence', flush=True)
oof_preds = np.zeros((len(train_features), len(target_columns)))
eval_loss_function = bce_loss
for fold_index, (train_index, val_index) in enumerate(skf.split(train_features, train_targets[y_labels])):
print(f'Infarence Fold: {fold_index}', train_index.shape, val_index.shape, flush=True)
X_valid = train_features.loc[val_index, all_features_columns].copy().values
y_valid = train_targets.iloc[val_index, 1:].copy().values
X_test = test_features[all_features_columns].values
if config.normalizer == 'log':
scaler = load_pickle(f'{model_dir}/scaler-{fold_index}.pkl')
X_valid = scaler.transform(X_valid)
X_test = scaler.transform(X_test)
transformer = load_pickle(f'{model_dir}/transformer-{fold_index}.pkl')
model = MoAEfficientNet(
pretrained_model_name=pretrained_model,
fc_size=config.fc_size,
drop_rate=config.drop_rate,
drop_connect_rate=config.drop_connect_rate,
weight_init='goog',
).to(DEVICE)
model.load_state_dict(torch.load(f'./{model_dir}/deepinsight-{fold_index}.pt'))
dataset = MoAImageDataset(X_valid, y_valid, transformer, image_size=config.image_size)
dataloader = DataLoader(
dataset,
batch_size=config.infer_batch_size,
shuffle=False,
num_workers=8,
pin_memory=True,
drop_last=False)
valid_loss, valid_preds = loop_valid(model, eval_loss_function, dataloader)
print(f'Fold {fold_index}/{config.kfolds} - fold_valid_loss: {valid_loss:.5f}', flush=True)
logger.update({'fold': fold_index, 'val_loss': valid_loss})
oof_preds[val_index, :] = valid_preds
dataset = TestDataset(X_test, None, transformer, image_size=config.image_size)
dataloader = DataLoader(
dataset,
batch_size=config.infer_batch_size,
shuffle=False,
num_workers=8,
pin_memory=True,
drop_last=False)
preds = loop_preds(model, dataloader)
test_preds += preds / config.kfolds
oof_preds_df = train_targets.copy()
oof_preds_df.loc[:, target_columns] = oof_preds.clip(0, 1)
oof_preds_df.to_csv(f'{model_dir}/oof_preds.csv', index=False)
oof_loss = mean_log_loss(train_targets.loc[:, target_columns].values, oof_preds)
print(f'OOF Validation Loss: {oof_loss:.6f}', flush=True)
print(f'Done infarence Elapsed {time_format(time.time() - start_time)}', flush=True)
logger.update({'fold': 'oof', 'val_loss': oof_loss})
logger.save(f'{model_dir}/log.csv')
submission = pd.DataFrame(data=test_features['sig_id'].values, columns=['sig_id'])
submission = submission.reindex(columns=['sig_id'] + target_columns)
submission.loc[:, target_columns] = test_preds.clip(0, 1)
submission.loc[test_features['cp_type'] == 0, submission.columns[1:]] = 0
submission.to_csv(f'{model_dir}/submission.csv', index=False)
def run(try_num, config):
logger = Logger()
args = get_args()
print('config:', config.to_dict(), flush=True)
print('args:', args, flush=True)
os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'
model_dir = f'blending-01-nn-{try_num}'
if not os.path.exists(model_dir):
os.mkdir(model_dir)
train_features = pd.read_csv('../input/lish-moa/train_features.csv')
train_targets = pd.read_csv('../input/lish-moa/train_targets_scored.csv')
dae_features = pd.read_csv(config.dae_path)
test_features = pd.read_csv('../input/lish-moa/test_features.csv')
if args.debug:
train_features = train_features[:500]
train_targets = train_targets[:500]
dae_features = pd.concat(
[dae_features.iloc[:500],
dae_features.iloc[-3982:]]).reset_index(drop=True)
config.update(
dict(
n_folds=3,
seeds=[222],
n_epochs=3,
batch_size=128,
))
target_columns = [col for col in train_targets.columns if col != 'sig_id']
n_targets = len(target_columns)
train_features, train_targets, test_features = preprocess(
config, model_dir, train_features, train_targets, test_features,
dae_features)
features_columns = [
col for col in train_features.columns if col not in [
'sig_id', 'cp_type', 'cp_time', 'cp_dose', 'cp_type_ctl_vehicle',
'cp_type_trt_cp'
]
]
metric_loss_function = nn.BCELoss()
if config.weighted_loss_strategy == 1:
indices = get_minority_target_index(
train_targets, threshold=config.weighted_loss_threshold)
indices = [int(i not in indices) for i, c in enumerate(target_columns)]
smooth_loss_function = SmoothBCELoss(
smoothing=config.smoothing,
weight=config.weighted_loss_weights,
weight_targets=indices,
n_labels=n_targets)
else:
smooth_loss_function = SmoothBCELoss(smoothing=config.smoothing)
kfold = MultilabelStratifiedKFold(n_splits=config.n_folds,
random_state=42,
shuffle=True)
for seed_index, seed in enumerate(config.seeds):
if args.only_pred:
print('Skip training', flush=True)
break
print(f'Train seed {seed}', flush=True)
set_seed(seed)
for fold_index, (train_indices, val_indices) in enumerate(
kfold.split(train_targets[target_columns].values,
train_targets[target_columns].values)):
print(f'Train fold {fold_index + 1}', flush=True)
x_train = train_features.loc[train_indices, features_columns]
y_train = train_targets.loc[train_indices, target_columns]
x_val = train_features.loc[val_indices, features_columns]
y_val = train_targets.loc[val_indices, target_columns]
model = new_model(config.model_kind,
len(features_columns)).to(DEVICE)
checkpoint_path = f'{model_dir}/repeat-{seed}_Fold-{fold_index + 1}.pt'
optimizer = optim.Adam(model.parameters(),
weight_decay=config.weight_decay,
lr=config.learning_rate)
scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer,
mode='min',
factor=0.1,
patience=3,
eps=1e-4,
verbose=True)
best_loss = np.inf
for epoch in range(config.n_epochs):
dataset = MoaDataset(x_train.values, y_train.values)
dataloader = DataLoader(dataset,
batch_size=config.batch_size,
shuffle=True,
drop_last=True)
train_loss = loop_train(model,
dataloader,
optimizer,
loss_functions=(
smooth_loss_function,
metric_loss_function,
))
dataset = MoaDataset(x_val.values, y_val.values)
dataloader = DataLoader(dataset,
batch_size=config.val_batch_size,
shuffle=False)
valid_loss, _ = loop_valid(model, dataloader,
metric_loss_function)
print(
'Epoch {}/{} - loss: {:5.5f} - val_loss: {:5.5f}'.
format(epoch + 1, config.n_epochs, train_loss, valid_loss),
flush=True)
logger.update({
'epoch': epoch + 1,
'loss': train_loss,
'val_loss': valid_loss
})
scheduler.step(valid_loss)
if valid_loss < best_loss:
best_loss = valid_loss
torch.save(model.state_dict(), checkpoint_path)
oof_preds = np.zeros((len(train_features), len(config.seeds), n_targets))
test_preds = np.zeros((len(test_features), n_targets))
for seed_index in range(len(config.seeds)):
seed = config.seeds[seed_index]
print(f'Inference for seed {seed}', flush=True)
_test_preds_in_seed = np.zeros((len(test_features), n_targets))
for fold_index, (_, valid_indices) in enumerate(
kfold.split(train_targets[target_columns].values,
train_targets[target_columns].values)):
x_val = train_features.loc[valid_indices, features_columns]
y_val = train_targets.loc[valid_indices, target_columns]
checkpoint_path = f'{model_dir}/repeat-{seed}_Fold-{fold_index + 1}.pt'
model = new_model(config.model_kind,
len(features_columns)).to(DEVICE)
model.load_state_dict(torch.load(checkpoint_path))
dataset = MoaDataset(x_val.values, y_val.values)
dataloader = DataLoader(dataset,
batch_size=config.val_batch_size,
shuffle=False)
preds = loop_pred(model, dataloader)
oof_preds[valid_indices, seed_index, :] = preds
dataset = MoaDataset(test_features[features_columns].values, None)
dataloader = DataLoader(dataset,
batch_size=config.val_batch_size,
shuffle=False)
preds = loop_pred(model, dataloader)
_test_preds_in_seed += preds / config.n_folds
score = mean_log_loss(train_targets.loc[:, target_columns].values,
oof_preds[:, seed_index, :],
n_targets=n_targets)
test_preds += _test_preds_in_seed / len(config.seeds)
print(f'Score for this seed {score:5.5f}', flush=True)
logger.update({'val_loss': score})
# Evalucate validation score
oof_preds = np.mean(oof_preds, axis=1)
score = mean_log_loss(train_targets.loc[:, target_columns].values,
oof_preds,
n_targets=n_targets)
print(f'Overall score is {score:5.5f}', flush=True)
# Save validation prediction
oof_pred_df = train_targets.copy()
oof_pred_df.iloc[:, 1:] = oof_preds
oof_pred_df.to_csv(f'{model_dir}/oof_pred.csv', index=False)
# Save log
logger.update({'val_loss': score})
logger.save(f'{model_dir}/log.csv')
# Save Test Prediction
test_features = pd.read_csv('../input/lish-moa/test_features.csv')
submission = create_submission(test_features, ['sig_id'] + target_columns)
submission[target_columns] = test_preds
submission.loc[test_features['cp_type'] == 'ctl_vehicle',
target_columns] = 0
submission.to_csv(f'{model_dir}/submission.csv', index=False)
#!/usr/bin/python # -*- coding: UTF-8 -*- """ # @Time : 2019/10/8 22:34 # @Author : peng.wang # @Email : [email protected] # @FileName: util.py # @ProjectName :sh-demand-forecast-alg """ import pandas as pd import numpy as np import requests import json import os from utils.misc import Logger log = Logger(log_path='log').logger class DataLoader(object): """ Data loader. Combines a dataset and a sampler, and provides an iterable over the given dataset. """ def __init__(self, data, train_len, pred_len, feature_names, target_name, append_train=False): self.data = data
