more

emodel_generalisation/adaptation.py

@@ -84,7 +84,6 @@ def build_resistance_models(
 
     # it seems dask does not quite work on this (to investigate, but multiprocessing is fast enough)
     df = func(df, access_point, parallel_factory="multiprocessing")
-
     models = {}
     for emodel in emodels:
         _df = df[df.emodel == emodel]
@@ -93,13 +92,16 @@ def build_resistance_models(
         if len(rin[rin < 0]) == 0:
             try:
                 coeffs, extra = Polynomial.fit(np.log10(scaler), np.log10(rin), 3, full=True)
-                if extra[0] < rcond_min:
-                    models[emodel] = {
-                        "resistance": {"polyfit_params": coeffs.convert().coef.tolist()},
-                        "shape": exemplar_data[key],
-                    }
+                if extra[0] > rcond_min:
+                    print(f"resistance fit for {key} of {emodel} is not so good")
+                models[emodel] = {
+                    "resistance": {"polyfit_params": coeffs.convert().coef.tolist()},
+                    "shape": exemplar_data[key],
+                }
             except (np.linalg.LinAlgError, TypeError):
                 print(f"fail to fit emodel {emodel}")
+        else:
+            print(f"resistance fit for {key} of {emodel} has negative rin")
     return df[df.emodel.isin(models)], models
 
 

emodel_generalisation/cli.py

@@ -755,8 +755,13 @@ def _get_resistance_models(exemplar_df, exemplar_data, scales_params):
         """We fit the scale/Rin relation for AIS and soma."""
         models = {}
         for emodel in exemplar_df.emodel:
+            Path(f"local/{emodel}").mkdir(parents=True, exist_ok=True)
             models[emodel] = build_all_resistance_models(
-                access_point, [emodel], exemplar_data[emodel], scales_params
+                access_point,
+                [emodel],
+                exemplar_data[emodel],
+                scales_params,
+                fig_path=Path(f"local/{emodel}"),
             )
         return models
 

emodel_generalisation/model/bpopt.py

@@ -832,7 +832,7 @@ def __init__(
         location,
         target_voltage=None,
         voltage_precision=0.1,
-        stimulus_duration=500.0,
+        stimulus_duration=1000.0,
         upper_bound=0.2,
         lower_bound=-0.2,
         strict_bounds=True,
@@ -912,7 +912,6 @@ def get_voltage_base(
         self, holding_current, cell_model, param_values, sim, isolate, timeout=None
     ):
         """Calculate voltage base for a certain holding current"""
-
         self.stimuli[0].amp = holding_current
         response = BPEMProtocol.run(
             self, cell_model, param_values, sim=sim, isolate=isolate, timeout=timeout
@@ -1102,7 +1101,6 @@ def __init__(
             "totduration": stimulus_totduration,
             "holding_current": None,
         }
-
         self.recording_name = f"{name}.{location.name}.v"
         stimulus = eCodes["step"](location=location, **stimulus_definition)
         recordings = [

emodel_generalisation/model/evaluation.py

@@ -451,7 +451,7 @@ def _define_Rin_protocol(
 
 
 def _define_holding_protocol(
-    efeatures, strict_bounds=False, ais_recording=False, max_depth=7, stimulus_duration=500.0
+    efeatures, strict_bounds=False, ais_recording=False, max_depth=7, stimulus_duration=1000.0
 ):
     """Define the search holding current protocol"""
     target_voltage = None
@@ -1013,7 +1013,7 @@ def __init__(
         self.rin_step_duration = 500.0
         self.rin_step_amp = -0.02
         self.rin_totduration = self.rin_step_delay + self.rin_step_duration
-        self.search_holding_duration = 500.0
+        self.search_holding_duration = 1000.0
         self.search_threshold_step_delay = 500.0
         self.search_threshold_step_duration = 2000.0
         self.search_threshold_totduration = 3000.0