Optimize datatypes of network states (#5)

* Optimize datatypes of network states

When simulation with no collective variable, network states were always saved in int64 or float64 which used unnecessary amount of memory.
numpy.min_scalar_type(num_opinions) determines the dtype.

sponet/cnvm/model.py

@@ -103,11 +103,13 @@ def simulate(
         if self.params.network_generator is not None:
             self.update_network()
 
+        opinion_dtype = np.min_scalar_type(self.params.num_opinions-1)
+
         if x_init is None:
             x_init = rng.choice(
                 np.arange(self.params.num_opinions), size=self.params.num_agents
             )
-        x = np.copy(x_init).astype(int)
+        x = np.copy(x_init).astype(opinion_dtype)
 
         t_delta = 0 if len_output is None else t_max / (len_output - 1)
 
@@ -142,7 +144,7 @@ def simulate(
             )
 
         t_traj = np.array(t_traj)
-        x_traj = np.array(x_traj, dtype=int)
+        x_traj = np.array(x_traj, dtype=opinion_dtype)
         if len_output is None:
             # remove duplicate subsequent states
             mask = mask_subsequent_duplicates(x_traj)

sponet/multiprocessing.py

@@ -151,8 +151,12 @@ def _sample_many_runs_subprocess(
     model = model_type(params)
 
     if collective_variable is None:
+
+        opinion_dtype = np.min_scalar_type(params.num_opinions-1)
+
         x_out = np.zeros(
-            (num_initial_states, num_runs, num_timesteps, model.params.num_agents)
+            (num_initial_states, num_runs, num_timesteps, model.params.num_agents),
+            dtype=opinion_dtype
         )
     else:
         x_out = np.zeros(

tests/test_multiprocessing.py

@@ -114,3 +114,29 @@ def test_parallelization_with_cv(self):
                 self.num_opinions,
             ),
         )
+
+    def test_output_dtype_no_cv(self):
+        num_opinions_list = [2, 257]
+        correct_dtype_list = [np.uint8, np.uint16]
+
+        for num_opinions, correct_dtype in zip(num_opinions_list, correct_dtype_list):
+
+            params = CNVMParameters(
+                num_opinions=num_opinions,
+                num_agents=self.num_agents,
+                r=1,
+                r_tilde=1,
+            )
+
+            t, x = sample_many_runs(
+                params=params,
+                initial_states=self.initial_states,
+                t_max=5,
+                num_timesteps=self.num_timesteps,
+                num_runs=2,
+                n_jobs=2,
+                collective_variable=None,
+            )
+
+            self.assertEqual(correct_dtype, x.dtype)
+

tests/tests_cnvm/test_model.py

@@ -36,6 +36,7 @@ def setUp(self):
             r_tilde=self.r_tilde,
         )
 
+
     def test_output(self):
         model = CNVM(self.params_complete)
         t_max = 100
@@ -97,3 +98,30 @@ def test_output_concise(self):
 
         for i in range(x.shape[0] - 1):
             self.assertFalse(np.allclose(x[i], x[i + 1]))
+
+    def test_output_dtype(self):
+
+        num_opinions_list = [2, 257]
+        correct_dtype_list = [np.uint8, np.uint16]
+
+        for num_opinions, correct_dtype in zip(num_opinions_list, correct_dtype_list):
+            params = CNVMParameters(
+                num_opinions=num_opinions,
+                num_agents=self.num_agents,
+                r=1,
+                r_tilde=1,
+            )
+
+            model = CNVM(params)
+            t_max = 5
+            t, x = model.simulate(t_max)
+
+            self.assertEqual(correct_dtype, x.dtype)
+
+
+
+
+
+
+
+