cb_test = cb.Pool(data.X_test, data.y_test)
if data.task == 'Ranking':
cb_test.set_group_id(data.groups)
preds = self.model.predict(cb_test)
metric = self.eval(data, preds)
return metric
def model_name(self):
name = "catboost_"
use_cpu = "gpu_" if self.use_gpu else "cpu_"
nr = str(self.num_rounds) + "_"
return name + use_cpu + nr + str(self.max_depth)
if __name__ == "__main__":
X, y, groups = du.get_yahoo()
dataset = Dataset(name='yahoo',
task='Ranking',
metric='NDCG',
get_func=du.get_yahoo)
print(dataset.X_train.shape)
print(dataset.y_test.shape)
t_start = time.time()
xgbModel = CatboostModel()
xgbModel.use_gpu = False
xgbModel.run_model(data=dataset)
eplased = time.time() - t_start
print("--------->> " + str(eplased))
def cifar():
return Dataset(name='cifar10',
task='Multiclass classification',
metric='Accuracy',
get_func=du.get_cifar10)
def news20():
return Dataset(name='news20',
task='Multiclass classification',
metric='Accuracy',
get_func=du.get_news20)
def log1p():
return Dataset(name='log1p',
task='Regression',
metric='RMSE',
get_func=du.get_log1p)
def higgs():
return Dataset(name='higgs',
task='Regression',
metric='RMSE',
get_func=du.get_higgs)
def __init__(self, data_func, name, task, metric):
self.data = Dataset(data_func())
self.name = name
self.task = task
self.metric = metric
def _predict(self, data):
pred = self.model.predict(data.X_test)
metric = self.eval(data, pred)
return metric
def model_name(self):
name = "thundergbm_"
use_cpu = "gpu_" if self.use_gpu else "cpu_"
nr = str(self.num_rounds) + "_"
return name + use_cpu + nr + str(self.max_depth)
if __name__ == "__main__":
# X, y = du.get_higgs()
dataset = Dataset(name='higgs',
task='Regression',
metric='RMSE',
get_func=du.get_realsim)
print(dataset.X_train.shape)
print(dataset.y_test.shape)
t_start = time.time()
tgmModel = ThunderGBMModel()
tgmModel.tree_method = 'hist'
tgmModel.run_model(data=dataset)
eplased = time.time() - t_start
print("--------->> " + str(eplased))
'Adam',
loss=sm.losses.bce_jaccard_loss,
metrics=metrics,
)
def test_preprocessing(image, mask):
sample = {}
image = cv2.resize(image, (cfg.image_size, cfg.image_size),
interpolation=cv2.INTER_NEAREST)
mask = cv2.resize(mask, (cfg.image_size, cfg.image_size),
interpolation=cv2.INTER_NEAREST)
sample["image"] = image
sample["mask"] = mask
# cv2.imshow("image", image)
# cv2.imshow("mask", mask)
# cv2.waitKey(0)
return sample
# Dataset for testation images
test_dataset = Dataset(TEST_IMAGES,
TEST_MASKS,
classes=cfg.CLASSES,
preprocessing=test_preprocessing)
test_dataloader = Dataloader(test_dataset, batch_size=1, shuffle=False)
scores = model.evaluate_generator(test_dataloader, verbose=1)
print("Loss: {:.5}".format(scores[0]))
for metric, value in zip(metrics, scores[1:]):
print("mean {}: {:.5}".format(metric.__name__, value))