TypeError: 'NoneType' object has no attribute '__getitem__'

[xinxliu]

Platform information:

• Python 2.7
• MXNet 1.4

Hi, I met the following error when I ran the train.py under insightface/recognition. I am sure I use the recommended dataset MS1M-ArcFace from the baidu source.

Traceback (most recent call last):
File "train.py", line 366, in
main()
File "train.py", line 363, in main
train_net(args)
File "train.py", line 231, in train_net
images_filter = config.data_images_filter,
File "/home/liuxx/work/insightface/recognition/image_iter.py", line 42, in init
header, _ = recordio.unpack(s)
File "/home/liuxx/anaconda2/envs/mxnet/lib/python2.7/site-packages/mxnet/recordio.py", line 416, in unpack
header = IRHeader(*struct.unpack(_IR_FORMAT, s[:_IR_SIZE]))
TypeError: 'NoneType' object has no attribute 'getitem'

Then I manually load the dataset:
>>> imgrec = mx.recordio.MXIndexedRecordIO('datasets/faces_emore/train.idx', 'datasets/faces_emore/train.rec', 'r')
>>> s = imgrec.read_idx(0)
>>>s is None
True

The result of s is None is True. Hence is there something wrong with the dataset?

[xinxliu]

Now I know what's wrong. Unzipping the .zip file under Ubuntu may cause CRC error when the zipped file is larger than 4GB. Unzipping it under Windows can solve this.