update the calculation of num_batches_per_epoch

the number of batches per epoch also depends on the number of gpus. it should be `num_batches_per_epoch = (cifar10.NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN / FLAGS.batch_size / FLAGS.num_gpus)`
cshuangyi82 · Nov 1, 2018 · c803290 · c803290
1 parent 82e783e
commit c803290
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/tutorials/image/cifar10/cifar10_multi_gpu_train.py b/tutorials/image/cifar10/cifar10_multi_gpu_train.py
@@ -147,7 +147,7 @@ def train():
 
     # Calculate the learning rate schedule.
     num_batches_per_epoch = (cifar10.NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN /
-                             FLAGS.batch_size)
+                             FLAGS.batch_size / FLAGS.num_gpus)
     decay_steps = int(num_batches_per_epoch * cifar10.NUM_EPOCHS_PER_DECAY)
 
     # Decay the learning rate exponentially based on the number of steps.