oserror: unable to synchronously open file (truncated file: eof = 2940928, sblock->base_addr = 0, stored_eof = 4107541)

[katherine-Z-kx]

Hello, I have a problem like the title. My environment: python=3.8.18, h5py=3.10.0，pytorch=1.9.0 and windows 10. When I trained the neural network, this error occurred every 43 epochs. I try to set numworks=0, or numworks=8, but this error still occurs. Can anyone help me? please.

[takluyver]

Unfortunately, that error message usually means that the file has got corrupted. I've no idea why it would occur in a regular pattern, though. You may need to start with pytorch to understand what it's doing with h5py - the numworks parameter means nothing to me, for instance.

[Ofir408]

@katherine-Z-kx I have the same error. How did you solve it?