def __call__(self):
# cvで画像を分ける
print("Training")
if self.get("recreate_cv"):
make_cv(
train_df=pd.read_csv(osp.join(self.ROOT, "input", self.raw_dirname, "train.csv")),
cv_type=self.get("cv"),
out_path=osp.join(self.ROOT, "src", "cvs"),
n_splits=self.get("n_splits"),
seeds=self.get("seeds"),
)
if self.get("train_flag"):
for seed in self.seeds: # train by seed
seed_everything(seed)
cv_df = pd.read_csv(osp.join(self.cv_path, f"{self.get('cv')}_{seed}.csv"))
for fold in self.get("run_folds"):
train_df = cv_df[(cv_df["fold"] != fold) & (cv_df["fold"] != -1)]
val_df = cv_df[cv_df["fold"] == fold]
self.train(train_df, val_df, seed, fold)
pos_rate = float(sum(target)) / target.shape[0]
logger.info('shape %s %s' % data.shape)
logger.info('pos num: %s, pos rate: %s' % (sum(target), pos_rate))
all_params = {
'max_depth': [9],
'n_estimators': [150],
'learning_rate': [0.1],
'scale_pos_weight': [1],
'min_child_weight': [0.01],
'subsample': [1],
'colsample_bytree': [0.5],
'reg_alpha': [0.01],
}
cv = make_cv(
) # cv = StratifiedKFold(target, n_folds=3, shuffle=True, random_state=0)
all_ans = None
all_target = None
all_ids = None
with open('train_feature_1.py', 'w') as f:
f.write("LIST_TRAIN_COL = ['" + "', '".join(feature_column) + "']\n\n")
logger.info('cv_start')
for params in ParameterGrid(all_params):
logger.info('param: %s' % (params))
for train_idx, test_idx in list(cv):
list_estimator = []
ans = []
insample_ans = []