Supposing that at the end of a simulation I find that more sweeps need to be performed, then is there a way to resume the simulation forward from the initially obtained ground state?
After reading some previous posts my guess is that it may be possible and we need to add
'MPS' : False and
'statics_initial' : <.bin file name here>
in the dictionary, but I am not completely sure about it and about the other changes we need to make.
It will be really helpful if can get some guidance in this matter.
Thanks!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
statics_initial is enough as a flag. The MPS to False probably came up in a
discussion about how to provide input state for the dynamics and skips the
ground state search completely, i.e., starting right away with the time
evolution.
Let's say you want to pick up the ground state search for the Ising model
from the Examples, copy-paste that file, change the write and output
directory, and add (constructed from the output directory, id, and unique
ID of the first simulation)
Then, you should be ready to go. If you compare the log files in the output
directory, you will find the IMPS output constructing the initial guess
without input file is now not present anymore.
Supposing that at the end of a simulation I find that more sweeps need to
be performed, then is there a way to resume the simulation forward from the
initially obtained ground state?
After reading some previous posts my guess is that it may be possible and
we need to add
'MPS' : False and
'statics_initial' : <.bin file name here>
in the dictionary, but I am not completely sure about it and about the
other changes we need to make.
It will be really helpful if can get some guidance in this matter.
P.S. Assuming that one added myObservables.AddObservable('State', True) in
the original examples, which stores states not only under a hash, but also
a human-readable filename.
statics_initial is enough as a flag. The MPS to False probably came up in a
discussion about how to provide input state for the dynamics and skips the
ground state search completely, i.e., starting right away with the time
evolution.
Let's say you want to pick up the ground state search for the Ising model
from the Examples, copy-paste that file, change the write and output
directory, and add (constructed from the output directory, id, and unique
ID of the first simulation)
Then, you should be ready to go. If you compare the log files in the output
directory, you will find the IMPS output constructing the initial guess
without input file is now not present anymore.
Bests
Daniel
On Fri, Dec 17, 2021 at 2:31 PM Amz Sin amzsin@users.sourceforge.net
wrote:
Dear Developers and Users,
Supposing that at the end of a simulation I find that more sweeps need to
be performed, then is there a way to resume the simulation forward from the
initially obtained ground state?
After reading some previous posts my guess is that it may be possible and
we need to add
'MPS' : False and
'statics_initial' : <.bin file name here>
in the dictionary, but I am not completely sure about it and about the
other changes we need to make.
It will be really helpful if can get some guidance in this matter.
Thanks!
Dear Developers and Users,
Supposing that at the end of a simulation I find that more sweeps need to be performed, then is there a way to resume the simulation forward from the initially obtained ground state?
After reading some previous posts my guess is that it may be possible and we need to add
'MPS' : False and
'statics_initial' : <.bin file name here>
in the dictionary, but I am not completely sure about it and about the other changes we need to make.
It will be really helpful if can get some guidance in this matter.
Thanks!
Hi,
statics_initial is enough as a flag. The MPS to False probably came up in a
discussion about how to provide input state for the dynamics and skips the
ground state search completely, i.e., starting right away with the time
evolution.
Let's say you want to pick up the ground state search for the Ising model
from the Examples, copy-paste that file, change the write and output
directory, and add (constructed from the output directory, id, and unique
ID of the first simulation)
str(g) + 'ObsOut0_1_State.bin',
Then, you should be ready to go. If you compare the log files in the output
directory, you will find the IMPS output constructing the initial guess
without input file is now not present anymore.
Bests
Daniel
On Fri, Dec 17, 2021 at 2:31 PM Amz Sin amzsin@users.sourceforge.net
wrote:
P.S. Assuming that one added myObservables.AddObservable('State', True) in
the original examples, which stores states not only under a hash, but also
a human-readable filename.
On Thu, Jan 6, 2022 at 10:37 AM Daniel Jaschke djaschke@users.sourceforge.net wrote: