increase checkpoint number

RLN_code/RLN_single_Train.py

@@ -18,7 +18,7 @@ import time
 base_dir = "/data"
 ground_truthdir = base_dir + '/train/ground_truth/'
 input_dir = base_dir + "/train/input/"
-
+test_image = base_dir + "/test/input/0.npz"
 learning_rate = 0.008
 timestr = time.strftime("%Y%m%d-%H%M%S")
 run_name = "/RLN_Final_Model_{}_{}/".format(learning_rate,timestr)
@@ -131,7 +131,7 @@ i = tf.Variable(0, trainable=False, dtype=tf.int64)
 
 #ckpt = tf.train.Checkpoint(step=i, model=model, optimizer=optimizer)
 ckpt = tf.train.Checkpoint(step=i, model=model)
-manager = tf.train.CheckpointManager(ckpt, train_model_path, max_to_keep=5)
+manager = tf.train.CheckpointManager(ckpt, train_model_path, max_to_keep=20)
 
 for epoch in range(0, epochs):
     logging.info("\nStart of epoch %d" % (epoch,))
@@ -181,5 +181,12 @@ for epoch in range(0, epochs):
                 save_path = manager.save()
                 logging.info("Saved checkpoint for step {}: {}".format(int(ckpt.step), save_path))
                 logging.info("loss {:1.2f}".format(loss_value.numpy()))
+
+            if int(i)%301 == 0 and test_image is not None:
+                x = [images.read_file(test_image)]
+                inference = model(x, training=False)
+                tf.summary.image("test input", x, i,max_outputs=1)
+                tf.summary.image("test output", inference[0], i,max_outputs=1)
+                tf.summary.image("test estimation", inference[1], i,max_outputs=1)
         # Log every 200 batches.
         logging.info("Step %s", int(i))

RLN_code/RLN_single_inference.py

@@ -37,7 +37,7 @@ model = RLN_model_simple(name="test")
 i = tf.Variable(0, trainable=False, dtype=tf.int64)
 
 ckpt = tf.train.Checkpoint(step=i, model=model)
-manager = tf.train.CheckpointManager(ckpt, train_model_path, max_to_keep=5)
+manager = tf.train.CheckpointManager(ckpt, train_model_path, max_to_keep=20)
 ckpt.restore(manager.latest_checkpoint)
 logging.info("restore sucessfull")
 a = os.listdir(input_dir)