prepping testing variable, add notes

dominoes.ipynb

@@ -103,7 +103,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 9,
+   "execution_count": 4,
    "id": "40c715a5",
    "metadata": {},
    "outputs": [],
@@ -123,7 +123,18 @@
     " # (but of course that could be because of differential sensitivity to temperature..., should test that directly)\n",
     "# now trying with 1 encoding layer. \n",
     "\n",
-    "# :)\n"
+    "# :)\n",
+    "\n",
+    "\n",
+    "# TODO: \n",
+    "# DOMINOES SEQUENCER Comparison of max to real reward:\n",
+    "# - Add target to batch (can do post-hoc, even if not requested)\n",
+    "# - Measure reward of target\n",
+    "# - Add a 2D vector comparing max and real reward for each batch element!!\n",
+    "\n",
+    "# TSP Distance Traveled:\n",
+    "# - Explicitly measure the distance traveled by the agent in the TSP task\n",
+    "# - Compare to Held-Karp Solution\n"
    ]
   },
   {

dominoes/datasets/dominoe_dataset.py

@@ -196,7 +196,15 @@ def create_training_variables(self, num_nets, **train_parameters):
 
     def save_training_variables(self, training_variables, epoch_state, **train_parameters):
         """dataset specific training variable storage"""
-        pass  # nothing to do (usually update training_variables in place)
+        pass  # nothing to do (update training_variables in place)
+
+    def create_testing_variables(self, num_nets, **test_parameters):
+        """dataset specific testing variable storage"""
+        return {}  # nothing here yet, but ready for it in the future
+
+    def save_testing_variables(self, testing_variables, epoch_state, **test_parameters):
+        """dataset specific testing variable storage"""
+        pass  # nothing to do (update testing_variables in place)
 
     @torch.no_grad()
     def generate_batch(self, train=True, device=None, **kwargs):

dominoes/datasets/tsp_dataset.py

@@ -102,6 +102,14 @@ def save_training_variables(self, training_variables, epoch_state, **train_param
         """dataset specific training variable storage"""
         pass  # nothing to do (usually update training_variables in place)
 
+    def create_testing_variables(self, num_nets, **test_parameters):
+        """dataset specific testing variable storage"""
+        return {}  # nothing here yet, but ready for it in the future
+
+    def save_testing_variables(self, testing_variables, epoch_state, **test_parameters):
+        """dataset specific testing variable storage"""
+        pass  # nothing to do (update testing_variables in place)
+
     def get_max_possible_output(self):
         """
         get the maximum possible output for the dataset