OOM Error running resnet model tensorflow - tensorflow

Question

I am running the resenet model on EC2 g2(NVIDIA GRID K520) instance and seeing a OOM Error. I have tried various combinations of removing the code that uses a GPU, prefixing CUDA_VISIBLE_DEVICES='0' and also reducing the batch_size to 64. I am still failing to start the training. Can you help?
W tensorflow/core/common_runtime/bfc_allocator.cc:270] **********************x***************************************************************************xx
W tensorflow/core/common_runtime/bfc_allocator.cc:271] Ran out of memory trying to allocate 196.00MiB. See logs for memory state.
W tensorflow/core/framework/op_kernel.cc:936] Resource exhausted: OOM when allocating tensor with shape[64,16,224,224]
E tensorflow/core/client/tensor_c_api.cc:485] OOM when allocating tensor with shape[64,16,224,224]
[[Node: unit_1_2/sub1/conv1/Conv2D = Conv2D[T=DT_FLOAT, data_format="NHWC", padding="SAME", strides=[1, 1, 1, 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/gpu:0"](unit_1_2/residual_only_activation/leaky_relu, unit_1_2/sub1/conv1/DW/read)]]
[[Node: train_step/update/_1561 = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/cpu:0", send_device="/job:localhost/replica:0/task:0/gpu:0", send_device_incarnation=1, tensor_name="edge_10115_train_step/update", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"]()]]
Traceback (most recent call last):
File "./resnet_main.py", line 203, in <module>
tf.app.run()
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 30, in run
sys.exit(main(sys.argv))
File "./resnet_main.py", line 197, in main
train(hps)
File "./resnet_main.py", line 82, in train
feed_dict={model.lrn_rate: lrn_rate})
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 382, in run
run_metadata_ptr)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 655, in _run
feed_dict_string, options, run_metadata)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 723, in _do_run
target_list, options, run_metadata)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 743, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors.ResourceExhaustedError: OOM when allocating tensor with shape[64,16,224,224]
[[Node: unit_1_2/sub1/conv1/Conv2D = Conv2D[T=DT_FLOAT, data_format="NHWC", padding="SAME", strides=[1, 1, 1, 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/gpu:0"](unit_1_2/residual_only_activation/leaky_relu, unit_1_2/sub1/conv1/DW/read)]]
[[Node: train_step/update/_1561 = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/cpu:0", send_device="/job:localhost/replica:0/task:0/gpu:0", send_device_incarnation=1, tensor_name="edge_10115_train_step/update", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"]()]]
Caused by op u'unit_1_2/sub1/conv1/Conv2D', defined at:
File "./resnet_main.py", line 203, in <module>
tf.app.run()
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 30, in run
sys.exit(main(sys.argv))
File "./resnet_main.py", line 197, in main
train(hps)
File "./resnet_main.py", line 64, in train
model.build_graph()
File "/home/ubuntu/indu/tf-benchmark/resnet/resnet_model.py", line 59, in build_graph
self._build_model()
File "/home/ubuntu/indu/tf-benchmark/resnet/resnet_model.py", line 94, in _build_model
x = res_func(x, filters[1], filters[1], self._stride_arr(1), False)
File "/home/ubuntu/indu/tf-benchmark/resnet/resnet_model.py", line 208, in _residual
x = self._conv('conv1', x, 3, in_filter, out_filter, stride)
File "/home/ubuntu/indu/tf-benchmark/resnet/resnet_model.py", line 279, in _conv
return tf.nn.conv2d(x, kernel, strides, padding='SAME')
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/gen_nn_ops.py", line 394, in conv2d
data_format=data_format, name=name)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/op_def_library.py", line 703, in apply_op
op_def=op_def)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 2310, in create_op
original_op=self._default_original_op, op_def=op_def)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 1232, in __init__
self._traceback = _extract_stack()

Answer 1

The NVIDIA GRID K520 has 8GB of memory (link). I have successfully trained ResNet models on a NVIDIA GPU with 12GB of memory. As the error suggests, TensorFlow attempts to put all of the networks weights in the GPU memory and fails. I believe you have a few options:
Train only on the CPU, as mentioned in the comments, assuming your CPU has more than 8GB of memory. This is not recommended.
Train a different network with fewer parameters. Several networks have been released since Resnet, such as Inception-v4, Inception-ResNet, with fewer parameters and comparable accuracy. This option costs nothing to try!
Buy a GPU with more memory. Easiest option if you have the money.
Buy another GPU with the same memory and train the bottom half of the network on one, and the top half of the network on the other. The difficulty in communicating between the GPUs makes this option less desirable.
I hope this helps you and others that run into similar memory issues.