tensorboard callback

iss/models/Callbacks.py

@@ -1,29 +1,77 @@
 # -*- coding: utf-8 -*-
 
-from keras.callbacks import Callback
+import os
+import datetime
 import numpy as np
-from iss.tools.tools import Tools
+import tensorflow as tf
+from keras.callbacks import Callback
 from IPython.display import display
 
+from iss.tools.tools import Tools
+
+
 class DisplayPictureCallback(Callback):
 
-	def __init__(self, model, epoch_laps, data_loader):
+    def __init__(self, model, epoch_laps, data_loader):
 
-		self.model_class = model
+        self.model_class = model
-		self.epoch_laps = epoch_laps
+        self.epoch_laps = epoch_laps
-		self.data_loader = data_loader
+        self.data_loader = data_loader
-		super(DisplayPictureCallback, self).__init__()
+        super(DisplayPictureCallback, self).__init__()
 
 
-	def on_epoch_end(self, epoch, logs):
+    def on_epoch_end(self, epoch, logs):
-		if epoch % self.epoch_laps == 0:
+        if epoch % self.epoch_laps == 0:
 
-			input_pict = self.data_loader.next()[0][1]
+            input_pict = self.data_loader.next()[0][1]
-			output_pict = self.model_class.predict_one(input_pict)
+            output_pict = self.model_class.predict_one(input_pict)
 
-			display(Tools.display_one_picture_scaled(input_pict))
+            display(Tools.display_one_picture_scaled(input_pict))
-			display(Tools.display_index_picture_scaled(output_pict))
+            display(Tools.display_index_picture_scaled(output_pict))
-		
+        
-		return self
+        return self
+
+class TensorboardCallback(Callback):
+
+    def __init__(self, log_dir, limit_image = 1, model = None, data_loader = None):
+        self.log_dir = os.path.join(log_dir, datetime.datetime.now().strftime("%Y%m%d-%H%M%S"))
+        self.limit_image = limit_image
+        self.model_class = model
+        self.data_loader = data_loader
+        self.writer = tf.summary.FileWriter(self.log_dir)
+        super(TensorboardCallback, self).__init__()
+
+    def on_epoch_end(self, epoch, logs=None):
+        print(logs)
+        image_summaries = []
+        
+        for input_pict in self.data_loader.next()[0][:self.limit_image]:
+            output_pict = self.model_class.predict_one(input_pict)[0]
+            input_im_bytes = Tools.bytes_image(input_pict*255)
+            output_im_bytes = Tools.bytes_image(output_pict*255)
+
+            image_summaries.append(tf.Summary.Value(tag = 'input', image = tf.Summary.Image(encoded_image_string = input_im_bytes)))
+            image_summaries.append(tf.Summary.Value(tag = 'output', image = tf.Summary.Image(encoded_image_string = output_im_bytes)))
+
+
+        image_summary = tf.Summary(value = image_summaries)
+        self.writer.add_summary(image_summary, epoch) 
+        self._write_logs(logs, epoch)
+
+        return self
+
+    def _write_logs(self, logs, index):
+        for name, value in logs.items():
+            if name in ['batch', 'size']:
+                continue
+            summary = tf.Summary()
+            summary_value = summary.value.add()
+            if isinstance(value, np.ndarray):
+                summary_value.simple_value = value.item()
+            else:
+                summary_value.simple_value = value
+            summary_value.tag = name
+            self.writer.add_summary(summary, index)
+        self.writer.flush()
 
 

iss/models/ModelTrainer.py

@@ -1,7 +1,7 @@
 # -*- coding: utf-8 -*-
 
 from keras.callbacks import ModelCheckpoint, CSVLogger
-from iss.models.Callbacks import DisplayPictureCallback
+from iss.models.Callbacks import DisplayPictureCallback, TensorboardCallback
 from iss.tools.tools import Tools
 import keras
 
@@ -91,12 +91,11 @@ class ModelTrainer:
 		if 'tensorboard' in config['callbacks']:
 			log_dir = config['callbacks']['tensorboard']['log_dir']
 			Tools.create_dir_if_not_exists(log_dir)
-			self.callbacks.extend([keras.callbacks.TensorBoard(
+			self.callbacks.extend([TensorboardCallback(
 				log_dir = log_dir,
-				histogram_freq=0,
+				limit_image = config['callbacks']['tensorboard']['limit_image'],
-				batch_size=32,
+				model = self.model,
-				write_graph=False,
+				data_loader = self.data_loader.get_train_generator()
-				write_images = True
 			)])