adapted estimator graph to handle no training

ie direct passing of closed form estimators

Author: davidsebfischer

batchglm/models/base/estimator.py

@@ -80,7 +80,7 @@ def loss(self):
 
     @property
     @abc.abstractmethod
-    def gradient(self):
+    def gradients(self):
         pass
 
 

batchglm/models/base_glm/estimator.py

@@ -12,7 +12,7 @@
 ESTIMATOR_PARAMS.update({
     "loss": (),
     "log_likelihood": ("features",),
-    "gradient": ("features",),
+    "gradients": ("features",),
     "hessians": ("features", "delta_var0", "delta_var1"),
     "fisher_inv": ("features", "delta_var0", "delta_var1"),
 })
@@ -32,8 +32,8 @@ def log_likelihood(self):
         return self.params["log_likelihood"]
 
     @property
-    def gradient(self):
-        return self.params["gradient"]
+    def gradients(self):
+        return self.params["gradients"]
 
     @property
     def hessians(self):

batchglm/models/glm_nb/estimator.py

@@ -23,7 +23,7 @@ def __init__(self, estim: AbstractEstimator):
         # causes evaluation of the properties that have not been computed during
         # training, such as the hessian.
         params = estim.to_xarray(
-            ["a_var", "b_var", "loss", "log_likelihood", "gradient", "hessians", "fisher_inv"],
+            ["a_var", "b_var", "loss", "log_likelihood", "gradients", "hessians", "fisher_inv"],
             coords=input_data.data
         )