Merge pull request #387 from leondavi/stat_update

[stat_update]

src_py/apiServer/stats.py

@@ -85,10 +85,25 @@ def get_loss_ts(self , plot : bool = False , saveToFile : bool = False):
         self.loss_ts_pd = df
 
         if plot:
+            sns.set(style="whitegrid")
+            plt.figure(figsize=(12, 8))
+            plt.gca().set_facecolor('lightblue')  # Set the background color to light blue
+
+            # Customize the grid lines to be black
+            plt.grid(color='black', linestyle='-', linewidth=0.5)
+
             sns.lineplot(data=df)
             plt.xlabel('Batch Num.')
             plt.ylabel('Loss Value')
-            plt.title('Training Loss Function')
+            plt.title(f'Training Loss Function of ({self.experiment_phase.get_name()})')
+
+            # Move legend outside of the plot
+            plt.legend(title="Worker", loc='center left', bbox_to_anchor=(1, 0.5), ncol=1)
+
+            if saveToFile:
+                plt.savefig('training_loss_function.png', bbox_inches='tight')
+
+            plt.show()
         return df
 
     def get_min_loss(self , plot : bool = False , saveToFile : bool = False): # Todo change it
@@ -331,6 +346,48 @@ def missed_batches_key(phase_name, source_name, worker_name):
                 break
         return missed_batches_dict
 
+    def plot_batches_status(self, plot=False):
+        workers_model_db_list = self.nerl_model_db.get_workers_model_db_list()
+        workers_names = [worker_model_db.get_worker_name() for worker_model_db in workers_model_db_list]
+        received_batches = self.get_recieved_batches()
+        missed_batches = self.get_missed_batches()
+
+        # Initialize dictionaries to store batch counts for each worker
+        batches_received_train = {worker: 0 for worker in workers_names}
+        batches_dropped_train = {worker: 0 for worker in workers_names}
+
+        # Fill the dictionaries with the counts of received and missed batches
+        for key, batches in received_batches.items():
+            worker = key.split('->')[-1]
+            batches_received_train[worker] += len(batches)
+
+        for key, batches in missed_batches.items():
+            worker = key.split('->')[-1]
+            batches_dropped_train[worker] += len(batches)
+
+        # Create a DataFrame for plotting
+        workers_comm_dict = {
+            'Worker': list(batches_received_train.keys()),
+            'batches_received_train': list(batches_received_train.values()),
+            'batches_dropped_train': list(batches_dropped_train.values())
+        }
+        df_train = pd.DataFrame(workers_comm_dict)
+
+        # Sort the DataFrame by the worker names
+        df_train = df_train.sort_values(by='Worker')
+
+        # Plotting
+        if plot:
+            plt.figure(figsize=(10, 6))
+            data_train = pd.melt(df_train, id_vars=['Worker'], value_vars=['batches_received_train', 'batches_dropped_train'])
+            batches_stats = sns.barplot(x='Worker', y='value', hue='variable', data=data_train, order=sorted(workers_names))
+            plt.ylabel('Number Of Batches')
+            plt.xlabel('Worker')
+            plt.title(f"Received & Dropped Batches At Freq. 5B/s ({self.experiment_phase.get_name()})")
+
+            batches_stats.legend(loc='upper right', bbox_to_anchor=(1.5, 0.2), shadow=True, ncol=1)
+            plt.show()       
+
     def get_communication_stats_workers(self):
         # return dictionary of {worker : {communication_stats}}
         communication_stats_workers_dict = OrderedDict()