hparams = {
'name': 'knrm',
'batch_size': 100,
'title_size': 20,
'his_size': 50,
'npratio': 4,
'embedding_dim': 300,
'kernel_num': 11,
'metrics': 'group_auc,ndcg@5,ndcg@10,mean_mrr',
'attrs': ['title'],
}
hparams = load_hparams(hparams)
device = torch.device(hparams['device'])
vocab, loader_train, loader_test, loader_validate = prepare(hparams,
validate=True)
knrmModel = KNRMModel(vocab=vocab, hparams=hparams).to(device)
if hparams['mode'] == 'test':
knrmModel.load_state_dict(torch.load(hparams['save_path']))
print("testing...")
evaluate(knrmModel, hparams, loader_test)
elif hparams['mode'] == 'train':
train(knrmModel,
hparams,
loader_train,
loader_test,
loader_validate,
tb=True)
from model.yolo_v1 import yolo_v1
import cv2
from utils import utils
from config import config_yolo_v1 as config
from sklearn.utils import shuffle
if __name__ == '__main__':
yolo = yolo_v1(True)
gt_images, gt_labels = utils.prepare()
print(len(gt_labels))
print(gt_images.shape)
print(gt_labels.shape)
# for step in range(0, self.max_step + 1):
for epoch in range(config.EPOCHS):
print('shuffle!!!!')
gt_images, gt_labels = shuffle(gt_images, gt_labels)
batches = utils.make_batches(gt_images, gt_labels)
for batch_idx, batch in enumerate(batches):
result = yolo.train(epoch, batch_idx, batch)
print(result)