Merge pull request #572 from sandialabs/feature-report-tweaks

HTML Report Updates

Author: coreyostrove

README.md

@@ -197,3 +197,29 @@ Questions?
 ----------
 For help and support with pyGSTi, please contact the authors at
 pygsti@sandia.gov.
+
+How To Cite pyGSTi
+------------------
+
+If you've used pyGSTi in the your research and are interested in citing
+us, please consider the following software design paper from some of the
+members of our development team (bibtex below):
+
+```
+@ARTICLE{Nielsen2020-rd,
+  title     = "Probing quantum processor performance with {py{GST}i}",
+  author    = "Nielsen, Erik and Rudinger, Kenneth and Proctor, Timothy and
+              Russo, Antonio and Young, Kevin and Blume-Kohout, Robin",
+  journal   = "Quantum Sci. Technol.",
+  publisher = "IOP Publishing",
+  volume    =  5,
+  number    =  4,
+  pages     = "044002",
+  month     =  jul,
+  year      =  2020,
+  url       = "https://iopscience.iop.org/article/10.1088/2058-9565/ab8aa4",
+  copyright = "http://iopscience.iop.org/page/copyright",
+  issn      = "2058-9565",
+  doi       = "10.1088/2058-9565/ab8aa4"
+}
+```

pygsti/algorithms/core.py

@@ -765,19 +765,21 @@ def run_iterative_gst(dataset, start_model, circuit_lists,
 
     models = []
     optimums = []
+    mdc_store_list = []
 
     for i in range(len(circuit_lists)):
         #then do the final iteration slightly differently since the generator should
         #give three return values.
         if i==len(circuit_lists)-1:
-            mdl_iter, opt_iter, final_objfn =  next(gst_iter_gen)
+                mdl_iter, opt_iter, mdc_store_iter, final_objfn = next(gst_iter_gen)
         else:
-            mdl_iter, opt_iter =  next(gst_iter_gen)
+            mdl_iter, opt_iter, mdc_store_iter =  next(gst_iter_gen)
 
         models.append(mdl_iter)
         optimums.append(opt_iter)
+        mdc_store_list.append(mdc_store_iter)
 
-    return models, optimums, final_objfn
+    return models, optimums, final_objfn, mdc_store_list
 
 def iterative_gst_generator(dataset, start_model, circuit_lists,
                       optimizer, iteration_objfn_builders, final_objfn_builders,
@@ -927,7 +929,6 @@ def _max_array_types(artypes_list):  # get the maximum number of each array type
                     first_iter_optimizer = _copy.deepcopy(optimizer)  # use a separate copy of optimizer, as it
                     first_iter_optimizer.fditer = optimizer.first_fditer  # is a persistent object (so don't modify!)
                     opt_result, mdc_store = run_gst_fit(mdc_store, first_iter_optimizer, obj_fn_builder, printer - 1)
-
                 else:
                     opt_result, mdc_store = run_gst_fit(mdc_store, optimizer, obj_fn_builder, printer - 1)
                 profiler.add_time('run_iterative_gst: iter %d %s-opt' % (i + 1, obj_fn_builder.name), tNxt)
@@ -944,7 +945,6 @@ def _max_array_types(artypes_list):  # get the maximum number of each array type
                     mdl.basis = start_model.basis
                     opt_result, mdc_store = run_gst_fit(mdc_store, optimizer, obj_fn_builder, printer - 1)
                     profiler.add_time('run_iterative_gst: final %s opt' % obj_fn_builder.name, tNxt)
-
                 tNxt = _time.time()
                 printer.log("Final optimization took %.1fs\n" % (tNxt - tRef), 2)
                 tRef = tNxt
@@ -954,11 +954,13 @@ def _max_array_types(artypes_list):  # get the maximum number of each array type
                 # Note: initial_mdc_store is *not* an objective fn (it's just a store) so don't send it back.
                 if mdc_store is not initial_mdc_store:
                     final_objfn = mdc_store
-
-                yield (mdc_store.model, opt_result, final_objfn)
+                yield (mdc_store.model, opt_result, mdc_store, final_objfn)
             else:
-                #If not the final iteration then only send back a copy of the model and the optimizer results
-                yield (mdc_store.model.copy(), opt_result)
+                #If not the final iteration then send back a copy of the model and the optimizer results
+                #mdc_store gets re-initialized at the start of each circuit list iteration, and doesn't appear
+                #to get propagated beyond that point so sending it back directly without copying (which would
+                #probably require implementing a custom method for MDC store objects) should be fairly safe.
+                yield (mdc_store.model.copy(), opt_result, mdc_store)
 
     printer.log('Iterative GST Total Time: %.1fs' % (_time.time() - tStart))
     profiler.add_time('run_iterative_gst: total time', tStart)

pygsti/algorithms/gaugeopt.py

@@ -265,6 +265,10 @@ def gaugeopt_custom(model, objective_fn, gauge_group=None,
     printer = _baseobjs.VerbosityPrinter.create_printer(verbosity, comm)
     tStart = _time.time()
 
+    #replace model with a new copy of itself so as to not propagate the conversion back to the 
+    #instance of the model object we are gauge optimizing.
+    model = model.copy()
+
     if comm is not None:
         mdl_cmp = comm.bcast(model if (comm.Get_rank() == 0) else None, root=0)
         try:

pygsti/circuits/circuitlist.py

@@ -225,8 +225,7 @@ def elementvec_to_array(self, elementvec, layout, mergeop="sum"):
             Dictates how to combine the `elementvec` components corresponding to a single
             plaquette entry (circuit).  If "sum", the returned array contains summed
             values.  If a format string, e.g. `"%.2f"`, then the so-formatted components
-            are joined together with separating commas, and the resulting array contains
-            string (object-type) entries.
+            are joined together, and the resulting array contains string (object-type) entries.
 
         Returns
         -------
@@ -241,7 +240,7 @@ def elementvec_to_array(self, elementvec, layout, mergeop="sum"):
             fmt = mergeop
             ret = _np.nan * _np.ones(len(self), dtype=_np.object_)
             for i,ckt in enumerate(self._circuits):
-                ret[i] = ", ".join(["NaN" if _np.isnan(x) else
+                ret[i] = "".join(["NaN" if _np.isnan(x) else
                                     (fmt % x) for x in elementvec[layout.indices(ckt)]])
         else:
             raise ValueError("Invalid `mergeop` arg: %s" % str(mergeop))

pygsti/circuits/circuitstructure.py

@@ -140,8 +140,7 @@ def elementvec_to_array(self, elementvec, layout, mergeop="sum"):
             Dictates how to combine the `elementvec` components corresponding to a single
             plaquette entry (circuit).  If "sum", the returned array contains summed
             values.  If a format string, e.g. `"%.2f"`, then the so-formatted components
-            are joined together with separating commas, and the resulting array contains
-            string (object-type) entries.
+            are joined, and the resulting array contains string (object-type) entries.
 
         Returns
         -------
@@ -155,7 +154,7 @@ def elementvec_to_array(self, elementvec, layout, mergeop="sum"):
             fmt = mergeop
             ret = _np.nan * _np.ones((self.num_rows, self.num_cols), dtype=_np.object_)
             for (i, j), opstr in self.elements.items():
-                ret[i, j] = ", ".join(["NaN" if _np.isnan(x) else
+                ret[i, j] = "".join(["NaN" if _np.isnan(x) else
                                        (fmt % x) for x in elementvec[layout.indices(opstr)]])
         else:
             raise ValueError("Invalid `mergeop` arg: %s" % str(mergeop))

pygsti/modelmembers/operations/lindbladerrorgen.py

@@ -259,15 +259,20 @@ def from_error_generator(cls, errgen_or_dim, parameterization="CPTPLND", element
             terms up to some order).  `"cterm"` is similar but uses Clifford operation
             action on stabilizer states.
 
-        state_space : `StateSpace` or castable to `StateSpace`
-            The state space upon which this error generator acts.
+        state_space : StateSpace, optional (default None)
+            StateSpace object to use in construction of this LindbladErrorgen.
+            If None we use the function `pygsti.baseobjs.statespace.default_space_for_dim`
+            to infer the correct state space from the dimensions of the passed in
+            error generator.
 
         Returns
         -------
         `LindbladErrorgen`
         """
-        errgen = _np.zeros((errgen_or_dim, errgen_or_dim), 'd') \
-            if isinstance(errgen_or_dim, (int, _np.int64)) else errgen_or_dim
+        if isinstance(errgen_or_dim, (int, _np.int64)):
+            errgen = _np.zeros((errgen_or_dim, errgen_or_dim), 'd')
+        else:
+            errgen =  errgen_or_dim
         return cls._from_error_generator(errgen, parameterization, elementary_errorgen_basis,
                                          mx_basis, truncate, evotype, state_space)
 
@@ -321,8 +326,11 @@ def from_error_generator_and_blocks(cls, errgen_or_dim, lindblad_coefficient_blo
             terms up to some order).  `"cterm"` is similar but uses Clifford operation
             action on stabilizer states.
 
-        state_space : `StateSpace` or castable to `StateSpace`
-            The state space upon which this error generator acts.
+        state_space : StateSpace, optional (default None)
+            StateSpace object to use in construction of this LindbladErrorgen.
+            If None we use the function `pygsti.baseobjs.statespace.default_space_for_dim`
+            to infer the correct state space from the dimensions of the passed in
+            error generator.
 
         Returns
         -------

pygsti/protocols/estimate.py

@@ -165,6 +165,8 @@ def __init__(self, parent, models=None, parameters=None, extra_parameters=None):
         self._final_objfn_cache = parameters.get('final_objfn_cache', None)
         self.final_objfn_builder = parameters.get('final_objfn_builder', _objfns.PoissonPicDeltaLogLFunction.builder())
         self._final_objfn = parameters.get('final_objfn', None)
+        self._per_iter_mdc_store = parameters.get('per_iter_mdc_store', None)
+
 
         self.extra_parameters = extra_parameters if (extra_parameters is not None) else {}
 
@@ -193,20 +195,22 @@ def __init__(self, parent, models=None, parameters=None, extra_parameters=None):
                               '_final_objfn_cache': 'dir-serialized-object',
                               'final_objfn_builder': 'serialized-object',
                               '_final_objfn': 'reset',
-                              '_gaugeopt_suite': 'serialized-object'
+                              '_gaugeopt_suite': 'serialized-object',
+                              '_per_iter_mdc_store': 'reset'
                               }
 
     @property
     def parameters(self):
         #HACK for now, until we can remove references that access these parameters
-        parameters = _collections.OrderedDict()
+        parameters = dict()
         parameters['protocol'] = self.protocol  # Estimates can hold sub-Protocols <=> sub-results
         parameters['profiler'] = self.profiler
         parameters['final_mdc_store'] = self._final_mdc_store
         parameters['final_objfn'] = self._final_objfn
         parameters['final_objfn_cache'] = self._final_objfn_cache
         parameters['final_objfn_builder'] = self.final_objfn_builder
         parameters['weights'] = self.circuit_weights
+        parameters['per_iter_mdc_store'] = self._per_iter_mdc_store
         parameters.update(self.extra_parameters)
         #parameters['raw_objective_values']
         #parameters['model_test_values']

pygsti/protocols/gst.py

@@ -1357,6 +1357,7 @@ def run(self, data, memlimit=None, comm=None, checkpoint=None, checkpoint_path=N
         if disable_checkpointing:
             seed_model = mdl_start.copy()
             mdl_lsgst_list = []
+            mdc_store_list = []
             starting_idx = 0
         else:
             # Set the checkpoint_path variable if None
@@ -1376,6 +1377,7 @@ def run(self, data, memlimit=None, comm=None, checkpoint=None, checkpoint_path=N
             if checkpoint is None:
                 seed_model = mdl_start.copy()
                 mdl_lsgst_list = []
+                mdc_store_list = []
                 checkpoint = GateSetTomographyCheckpoint()
             elif isinstance(checkpoint, GateSetTomographyCheckpoint):
                 # if the checkpoint's last completed iteration is non-negative
@@ -1393,8 +1395,12 @@ def run(self, data, memlimit=None, comm=None, checkpoint=None, checkpoint_path=N
                 # left set to None. There looks to be some logic for handling this and it looks
                 # like the serialization routines effectively do this already, as the value
                 # of this is lost between writing and reading.
+                mdc_store_list = [None]*len(mdl_lsgst_list) #We don't presently have serialization support for
+                #MDC store objects, so for now we'll skip serializing this and re-initialize previous iterations
+                #to None. Given how this is currently used the only downside to this should be inefficiency
+                #rebuilding the needed MDC stores in the report generation. 
             else:
-                NotImplementedError(
+                raise NotImplementedError(
                     'The only currently valid checkpoint inputs are None and GateSetTomographyCheckpoint.')
 
             # note the last_completed_iter value is initialized to -1 so the below line
@@ -1420,10 +1426,11 @@ def run(self, data, memlimit=None, comm=None, checkpoint=None, checkpoint_path=N
             #then do the final iteration slightly differently since the generator should
             #give three return values.
             if i==len(bulk_circuit_lists)-1:
-                mdl_iter, opt_iter, final_objfn = next(gst_iter_generator)
+                mdl_iter, opt_iter, mdc_store_iter, final_objfn = next(gst_iter_generator)
             else:
-                mdl_iter, opt_iter =  next(gst_iter_generator)
+                mdl_iter, opt_iter, mdc_store_iter =  next(gst_iter_generator)
             mdl_lsgst_list.append(mdl_iter)
+            mdc_store_list.append(mdc_store_iter)
             optima_list.append(opt_iter)
 
             if not disable_checkpointing:
@@ -1436,15 +1443,15 @@ def run(self, data, memlimit=None, comm=None, checkpoint=None, checkpoint_path=N
                     checkpoint.write(f'{checkpoint_path}_iteration_{i}.json')
 
         tnxt = _time.time(); profiler.add_time('GST: total iterative optimization', tref); tref = tnxt
-    
         #set parameters
-        parameters = _collections.OrderedDict()
+        parameters = dict()
         parameters['protocol'] = self  # Estimates can hold sub-Protocols <=> sub-results
         parameters['final_objfn_builder'] = self.objfn_builders.final_builders[-1] \
             if len(self.objfn_builders.final_builders) > 0 else self.objfn_builders.iteration_builders[-1]
         parameters['final_objfn'] = final_objfn  # Final obj. function evaluated at best-fit point (cache too)
         parameters['final_mdc_store'] = final_objfn  # Final obj. function is also a "MDC store"
         parameters['profiler'] = profiler
+        parameters['per_iter_mdc_store'] = mdc_store_list #list of the MDC stores for each iteration.
         # Note: we associate 'final_cache' with the Estimate, which means we assume that *all*
         # of the models in the estimate can use same evaltree, have the same default prep/POVMs, etc.