TensorFlow GPU not working in Nano

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

2019-09-23 21:47:33.538782: W tensorflow/core/common_runtime/bfc_allocator.cc:237] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.05GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-09-23 21:47:34.492977: W tensorflow/core/common_runtime/bfc_allocator.cc:237] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.27GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-09-23 21:47:34.573111: W tensorflow/core/common_runtime/bfc_allocator.cc:237] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.55GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-09-23 21:47:34.608569: W tensorflow/core/common_runtime/bfc_allocator.cc:237] Allocator (GPU_0_bfc) ran out of memory trying to allocate 804.75MiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-09-23 21:47:34.652985: W tensorflow/core/common_runtime/bfc_allocator.cc:237] Allocator (GPU_0_bfc) ran out of memory trying to allocate 3.06GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-09-23 21:47:34.853834: W tensorflow/core/common_runtime/bfc_allocator.cc:237] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.05GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-09-23 21:47:35.411436: W tensorflow/core/common_runtime/bfc_allocator.cc:237] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.06GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-09-23 21:47:35.471115: W tensorflow/core/common_runtime/bfc_allocator.cc:237] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.58GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-09-23 21:47:35.976014: W tensorflow/core/common_runtime/bfc_allocator.cc:237] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.76GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-09-23 21:47:35.993860: W tensorflow/core/common_runtime/bfc_allocator.cc:237]Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.18GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
Gtk-Message: 21:47:46.457: Failed to load module “canberra-gtk-module”

instead of GPU it runs in CPU… can you give solution to this

2

What makes you think that GPU is not used? By the way, what are you trying to achieve - which software and model are you using?

3

Hi,

This is a known issue.

TensorFlow by default allocates all the available memory when creating session.
Such allocation may fail in the shared memory system, like Jetson.

You can set the allow_growth flag to avoid big chunk allocation.

cnfig = tf.ConfigProto()
config.gpu_options.allow_growth = True
config.gpu_options.per_process_gpu_memory_fraction = 0.4
session = tf.Session(config=config, ...)

However, please noticed that Nano only have 4G memory.
It looks like your use case take too much memory to run on Nano.

Thanks.