Dear there,
I met a strange problem which sometime happens but sometimes not. It’s when I try to read a previous stored MPS (a h5 file), sometime I got error message saying this:
Because I ususally run several jobs with same code just different parameter values at same time, and I always find some of them runs smoothly while some of them got this error, so I’m a bit confused. I’m using Julia/1.8.4, Itensor v0.6.16 and HDF5 v0.16.13 (I just tried update to HDF5 v0.17.2 but still same problem).
Look forward to your help! Thank you so much!
Sorry to hear you’re running into this problem. But it’s hard for us to know what might be causing it just from the information provided. Would it be possible to share a short piece of code that reproduces this error?
This looks to be an HDF5 issue with the internal library. My guess from there error is that there are multiple versions of libhdf5 installed on your machine and occasionally the wrong version is library linked when starting Julia.
Are you on a cluster module system?
I have found a much easier time saving and loading ITensor objects using the JLD2.jl package. So potentially this package might do a more robust job handling HDF5 versions?
Thanks for your reply! But I don’t have a code that I know for sure will generate a problematic h5 file becasue using same code, it just sometimes geenrates a good h5 sometimes generate a bad h5… I also suspect it may be related to something in the cluster I’m using so I’ll contact them also.
Or is there anyway I can upload a problematic h5 file by any chance…?
