Update model.py

remove extraneous layer names, abbreviate kernel divergence function name
tr7200 · Jan 30, 2021 · f6b6344 · f6b6344
1 parent cddd728
commit f6b6344
Showing 1 changed file with 5 additions and 5 deletions.
diff --git a/model.py b/model.py
@@ -51,15 +51,15 @@
     bd_input = Input(shape=(39,))
     as_input = Input(shape=(12,))
 
-    kernel_divergence_fn = lambda q, p, _: tfp.distributions.kl_divergence(q, p) / (323 * 1.0)
+    krnl_dvrgnc_fn = lambda q, p, _: tfp.distributions.kl_divergence(q, p) / (323 * 1.0)
 
 
     # outer chevron
     sm_bd_combined = concatenate([sm_input, bd_input])
-    sm_bd_combined_out = tfp.layers.DenseFlipout(48, activation='relu', name = 'dense1', kernel_divergence_fn=kernel_divergence_fn)(sm_bd_combined )
+    sm_bd_combined_out = tfp.layers.DenseFlipout(48, activation='relu', kernel_divergence_fn=krnl_dvrgnc_fn)(sm_bd_combined)
 
     as_and_sm_bd_combined = concatenate([sm_bd_combined_out, as_input])
-    as_and_sm_bd_combined_out = tfp.layers.DenseFlipout(3, activation='relu', name = 'dense2', kernel_divergence_fn=kernel_divergence_fn)(as_and_sm_bd_combined)
+    as_and_sm_bd_combined_out = tfp.layers.DenseFlipout(3, activation='relu', kernel_divergence_fn=krnl_dvrgnc_fn)(as_and_sm_bd_combined)
     V_struct_1_out  = tfp.layers.DistributionLambda(lambda t: tfp.distributions.Normal(loc=25 + t[..., :1],
                                                     validate_args=True,
                                                     allow_nan_stats=False,
@@ -78,8 +78,8 @@
     # inner chevron
     V1_out = V_struct_1([sm_input, bd_input, as_input])
     V1_SM_BD_combined = concatenate([V1_out, bd_input, sm_input])
-    V1_SM_BD_combined_out1 = tfp.layers.DenseFlipout(37, activation='relu', name = 'dense3', kernel_divergence_fn=kernel_divergence_fn)(V1_SM_BD_combined)
-    V1_SM_BD_combined_out2 = tfp.layers.DenseFlipout(10, activation='relu', name = 'dense4', kernel_divergence_fn=kernel_divergence_fn)(V1_SM_BD_combined_out1)
+    V1_SM_BD_combined_out1 = tfp.layers.DenseFlipout(37, activation='relu', kernel_divergence_fn=krnl_dvrgnc_fn)(V1_SM_BD_combined)
+    V1_SM_BD_combined_out2 = tfp.layers.DenseFlipout(10, activation='relu', kernel_divergence_fn=krnl_dvrgnc_fn)(V1_SM_BD_combined_out1)
     V2_out = tfp.layers.DistributionLambda(lambda t: tfp.distributions.Normal(loc=25 + t[..., :1], 
                                            validate_args=True, 
                                            allow_nan_stats=False,