Johnnie Gray

You can also use a autograd optimizer directly like so, this is a much more stable algorithm than periodic DMRG: ```python mps = qtn.MPS_rand_state(L, bond_dim=20, phys_dim=2, cyclic=True) def norm_fn(mps): return...

Regarding `TNOptimizer`, it should match... I can't replicate that behavior here https://colab.research.google.com/drive/1DgZrGizDACL1C2vDSx_wMxvr01PnrzLi?usp=sharing. Perhaps the optimizer is defaulting to using `jax` which defaults to single precision and there is some conditioning...

Thanks for catching this @kevincsmith, yes it would be great if you could submit a PR! Ideally with the test case, then we can quickly release a new version.

Thanks for the issue @thibxlv, this should be fixed by https://github.com/jcmgray/quimb/commit/fec27eb3d2801bd8268762e31f1f9c95d871cfe0. You should also be able to supply `backend_random="jax"` now to do the sampling itself in jax and so e.g....

Hi @projekter, thanks for the issue. Yes this might be similar to #32, i.e. on my list of to-dos for a long time but no-one had run into it until...

Hi @ermalrrapaj. Yes currently parametrized gates I think are not compatible with the MPS simulator like this, since it must eagerly contract in actual data arrays. It should be possible...

Hi @sajjan02purdue, I don't see anything obvious but sadly its too long an example for me to process what really what is going on or understand. I would recommend trying...

Hi @AndrewArrasmith, yes that compress call was added assuming calling classes might have an obvious default `compress` method, but outside of 1D and tree like networks there's not really one...

This is a bit confusing but intentional, the way indices works means that whether something has been 'transposed' in the matrix sense depends on how you compose it with other...

Hi @HomoY, I'm afraid I don't really have any ideas nor any way to debug this. Leaving open in case others have a similar problem though.