Unexpected bond dimension #126
-
I find that I am looking at the SU(2) t-J model on a ladder, using the example in https://block2.readthedocs.io/en/latest/tutorial/custom-hamiltonians.html#SU(2)-t-J-Model and the following parameters:
I get a ground-state energy If I use the modified symmetry type suggested here: #115 (comment) (required for converting to SZ), I get an even higher bond dimension for the same parameters, with The mismatch persists for nonzero How do I understand this? Is there a way to track the actual (maximum) bond dimension at the end of each sweep? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 3 replies
-
Thanks for reporting this problem. The value returned from As a workaround, you can add an argument This will be fixed in version |
Beta Was this translation helpful? Give feedback.
Thanks for reporting this problem. The value returned from
mps.info.get_max_bond_dimension()
is incorrect.As a workaround, you can add an argument
full_fci=False
todriver.get_random_mps
, thenmps.info.get_max_bond_dimension()
will behave correctly.This will be fixed in version
0.5.3rc19
so thatmps.info.get_max_bond_dimension()
will print the correct value with or withoutfull_fci=False
.