chenjw

一个小问题

1

``` # 还不懂啥意思 test_transformed['det_hh_fam_stat_ Grandchild

请问下，get_masks_and_position_ids 这个函数里面， attention_mask = torch.ones((1, context_length, context_length), device=device) 为什么 mask的shape 是 （1，context_length, context_length)呢？ 之前只用过bert，attention mask 都是tokenizer自己返回的。 这里的实现原理从哪可以学习下嘛？ 感谢～

` sub['week3'] = y_pred * 1.6 sub['week1'] = sub['week3'].map(lambda x: (x / 1.6) * 1) sub['week2'] = sub['week3'].map(lambda x: (x / 1.6) * 1.3) sub['week3'] = sub['week3'].map(lambda x: (x /...

season1里面代码问题

1

0.data4MLM_raw.py ``` num = 0 for line in tqdm(train_query): if num % 20 == 0: num += 1 writer.write(line) writer.write('\n') ``` num += 1 是不是要写在if外面