Merge branch 'master' of gitlab.doc.ic.ac.uk:yw21218/metarl

MetaAugment/MetaAugment/test/cifar-10-batches-py/readme.html

+<meta HTTP-EQUIV="REFRESH" content="0; url=http://www.cs.toronto.edu/~kriz/cifar.html">

MetaAugment/MetaAugment/train/cifar-10-batches-py/readme.html

+<meta HTTP-EQUIV="REFRESH" content="0; url=http://www.cs.toronto.edu/~kriz/cifar.html">

MetaAugment/autoaugment_learners/evo_learner.py

@@ -29,7 +29,6 @@ class evo_learner():
                 batch_size=8,
                 toy_flag=False,
                 toy_size=0.1,
-                sub_num_pol=5, 
                 fun_num = 14,
                 exclude_method=[],
                 ):
@@ -46,15 +45,15 @@ class evo_learner():
             max_epochs=max_epochs,
             early_stop_num=early_stop_num,)
 
-
-        self.auto_aug_agent = Evo_learner(fun_num=fun_num, p_bins=p_bins, m_bins=m_bins, sub_num_pol=sub_num_pol)
+        self.num_solutions = num_solutions
+        self.auto_aug_agent = Evo_learner(fun_num=fun_num, p_bins=p_bins, m_bins=m_bins, sub_num_pol=sp_num)
         self.torch_ga = torchga.TorchGA(model=self.auto_aug_agent, num_solutions=num_solutions)
         self.num_parents_mating = num_parents_mating
         self.initial_population = self.torch_ga.population_weights
         self.train_loader = train_loader
         self.child_network = child_network
         self.p_bins = p_bins 
-        self.sub_num_pol = sub_num_pol
+        self.sub_num_pol = sp_num
         self.m_bins = m_bins
         self.fun_num = fun_num
         self.augmentation_space = [x for x in augmentation_space if x[0] not in exclude_method]
@@ -121,15 +120,15 @@ class evo_learner():
         """
         section = self.auto_aug_agent.fun_num + self.auto_aug_agent.p_bins + self.auto_aug_agent.m_bins
 
-        y = self.auto_aug_agent.forward(x) # 1000 x 32
+        y = self.auto_aug_agent.forward(x)
 
-        y_1 = torch.softmax(y[:,:self.auto_aug_agent.fun_num], dim = 1) # 1000 x 14
+        y_1 = torch.softmax(y[:,:self.auto_aug_agent.fun_num], dim = 1) 
         y[:,:self.auto_aug_agent.fun_num] = y_1
         y_2 = torch.softmax(y[:,section:section+self.auto_aug_agent.fun_num], dim = 1)
         y[:,section:section+self.auto_aug_agent.fun_num] = y_2
         concat = torch.cat((y_1, y_2), dim = 1)
 
-        cov_mat = torch.cov(concat.T)#[:self.auto_aug_agent.fun_num, self.auto_aug_agent.fun_num:]
+        cov_mat = torch.cov(concat.T)
         cov_mat = cov_mat[:self.auto_aug_agent.fun_num, self.auto_aug_agent.fun_num:]
         shape_store = cov_mat.shape
 
@@ -197,9 +196,16 @@ class evo_learner():
             Solution_idx -> Int
         """
         self.num_generations = iterations
+        self.history_best = [0 for i in range(iterations)]
+        self.history_avg = [0 for i in range(iterations)]
+        self.gen_count = 0
+        self.best_model = 0
+
         self.set_up_instance()
 
         self.ga_instance.run()
+        self.history_avg = self.history_avg / self.num_solutions
+
         solution, solution_fitness, solution_idx = self.ga_instance.best_solution()
         if return_weights:
             return torchga.model_weights_as_dict(model=self.auto_aug_agent, weights_vector=solution)
@@ -207,14 +213,6 @@ class evo_learner():
             return solution, solution_fitness, solution_idx
 
 
-    def new_model(self):
-        """
-        Simple function to create a copy of the secondary model (used for classification)
-        """
-        copy_model = copy.deepcopy(self.child_network)
-        return copy_model
-
-
     def set_up_instance(self, train_dataset, test_dataset):
         """
         Initialises GA instance, as well as fitness and on_generation functions
@@ -249,9 +247,16 @@ class evo_learner():
                     full_policy = self.get_full_policy(test_x)
 
 
-            fit_val = ((self.test_autoaugment_policy(full_policy, train_dataset, test_dataset)[0])/
+            fit_val = ((self.test_autoaugment_policy(full_policy, train_dataset, test_dataset)[0]) /
                         + self.test_autoaugment_policy(full_policy, train_dataset, test_dataset)[0]) / 2
 
+            if fit_val > self.history_best[self.gen_count]:
+                self.history_best[self.gen_count] = fit_val 
+                self.best_model = model_weights_dict
+            
+            self.history_avg[self.gen_count] += fit_val
+            
+
             return fit_val
 
         def on_generation(ga_instance):
@@ -267,6 +272,7 @@ class evo_learner():
             None
             """
             print("Generation = {generation}".format(generation=ga_instance.generations_completed))
+            self.gen_count += 1
             print("Fitness    = {fitness}".format(fitness=ga_instance.best_solution()[1]))
             return