fixed bug that convergence isnt set to False after training_strategy is done

Author: davidsebfischer

batchglm/train/tf/base_glm_all/estimator.py

@@ -333,14 +333,15 @@ def train_sequence(self, training_strategy):
         if isinstance(training_strategy, Enum):
             training_strategy = training_strategy.value
         elif isinstance(training_strategy, str):
-            training_strategy = self.TrainingStrategy[training_strategy].value
+            training_strategy = self.TrainingStrategies[training_strategy].value
 
         if training_strategy is None:
-            training_strategy = self.TrainingStrategy.DEFAULT.value
+            training_strategy = self.TrainingStrategies.DEFAULT.value
 
         logger.info("training strategy:\n%s", pprint.pformat(training_strategy))
 
         for idx, d in enumerate(training_strategy):
+            self.model.model_vars.converged = False
             logger.info("Beginning with training sequence #%d", idx + 1)
             self.train(**d)
             logger.info("Training sequence #%d complete", idx + 1)

batchglm/train/tf/glm_nb/estimator.py

@@ -1,4 +1,3 @@
-from enum import Enum
 import logging
 from typing import Union
 
@@ -10,6 +9,7 @@
 from .external import closedform_nb_glm_logmu, closedform_nb_glm_logphi
 from .estimator_graph import EstimatorGraph
 from .model import ProcessModel
+from .training_strategies import TrainingStrategies
 
 logger = logging.getLogger(__name__)
 
@@ -20,33 +20,6 @@ class Estimator(EstimatorAll, AbstractEstimator, ProcessModel):
     Uses the natural logarithm as linker function.
     """
 
-    class TrainingStrategy(Enum):
-        AUTO = None
-        DEFAULT = [
-            {
-                "convergence_criteria": "all_converged_ll",
-                "stopping_criteria": 1e-6,
-                "use_batching": False,
-                "optim_algo": "irls",
-            },
-        ]
-        QUICK = [
-            {
-                "convergence_criteria": "all_converged_ll",
-                "stopping_criteria": 1e-4,
-                "use_batching": False,
-                "optim_algo": "irls",
-            },
-        ]
-        EXACT = [
-            {
-                "convergence_criteria": "all_converged_ll",
-                "stopping_criteria": 1e-8,
-                "use_batching": False,
-                "optim_algo": "irls",
-            },
-        ]
-
     def __init__(
             self,
             input_data: InputData,
@@ -62,6 +35,7 @@ def __init__(
             extended_summary=False,
             dtype="float64",
     ):
+        self.TrainingStrategies = TrainingStrategies
         EstimatorAll.__init__(
             self=self,
             input_data=input_data,

batchglm/train/tf/glm_nb/training_strategies.py

@@ -0,0 +1,43 @@
+from enum import Enum
+
+class TrainingStrategies(Enum):
+
+    AUTO = None
+    DEFAULT = [
+        {
+            "convergence_criteria": "all_converged_ll",
+            "stopping_criteria": 1e-6,
+            "use_batching": False,
+            "optim_algo": "irls",
+        },
+    ]
+    QUICK = [
+        {
+            "convergence_criteria": "all_converged_ll",
+            "stopping_criteria": 1e-3,
+            "use_batching": True,
+            "optim_algo": "irls",
+        },
+        {
+            "convergence_criteria": "all_converged_ll",
+            "stopping_criteria": 1e-6,
+            "use_batching": False,
+            "optim_algo": "irls",
+        },
+    ]
+    INEXACT = [
+        {
+            "convergence_criteria": "all_converged_ll",
+            "stopping_criteria": 1e-4,
+            "use_batching": False,
+            "optim_algo": "irls",
+        },
+    ]
+    EXACT = [
+        {
+            "convergence_criteria": "all_converged_ll",
+            "stopping_criteria": 1e-8,
+            "use_batching": False,
+            "optim_algo": "irls",
+        },
+    ]

batchglm/unit_test/glm_all/test_acc_glm_all.py

@@ -28,7 +28,7 @@ def __init__(
             else:
                 raise ValueError("noise_model not recognized")
 
-        batch_size = 900
+        batch_size = 200
         provide_optimizers = {"gd": True, "adam": True, "adagrad": True, "rmsprop": True, "nr": True, "irls": True}
         estimator = Estimator(
             input_data=simulator.input_data,