"python logbert.py train" fails with message "requested array has an inhomogeneous shape after 1 dimensions"

[kodefoundry]

% python logbert.py train
device cpu
features logkey:True time: False

mask ratio 0.65
arguments Namespace(mode='train')
Save options parameters
Loading vocab ../output/hdfs/vocab.pkl
vocab Size: 20

Loading Train Dataset
before filtering short session
train size 4370
valid size 485

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 4855/4855 [00:00<00:00, 110354.29it/s]
Traceback (most recent call last):
File "/workspace/datascience/logbert/logbert/HDFS/logbert.py", line 102, in
Trainer(options).train()
File "/workspace/datascience/logbert/logbert/bert_pytorch/train_log.py", line 62, in train
logkey_train, logkey_valid, time_train, time_valid = generate_train_valid(self.output_path + "train", window_size=self.window_size,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/workspace/datascience/logbert/logbert/bert_pytorch/dataset/sample.py", line 90, in generate_train_valid
logkey_seq_pairs = np.array(logkey_seq_pairs)
^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (4855,) + inhomogeneous part.

[yebo852058]

你在sample.py里面修改一下这个就好了，numpy版本问题。