chunk = read(handle, remaining) about 6-pack

[LyuJ1998]

Hi, I have fix the bug you mentioned in another issue. Check it and your problem may disappear. If not, please make sure again you have remove all of the "try-excepy" code, especially the line 488 and line 471 in dataset/dataset_nocs.py.

[sunshantong]

@LyuJ1998 Hi, thank you for your reply and update. The dataloader bug has been fixed. But I still have the same problem when I run the code. The code will still stick at testdataloader. When I set num_workers = 0, the code can run without multithreading. But it's very slow. Is it possible deadlock in dataloader? Thank you very much.

[sunshantong]

I fix the problem by calling the main script by OMP_NUM_THREADS=1 MKL_NUM_THREADS=1 python train.py Hope to be helpful to other similar situations.

[mystorm16]

Looks like this was a bug in Python
i fix this issue by changing my python version ：3.6.0 to 3.6.7
https://stackoverflow.com/questions/53300965/pytorch-exception-in-thread-valueerror-signal-number-32-out-of-range