TensorFlow : Docker イメージ (GPU) を利用する2020/07/24

	機械学習ライブラリー, TensorFlow をインストールします。当例では、TensorFlow 公式の Docker イメージをダウンロードして、コンテナーから TensorFlow を利用します。 Docker イメージは GPU サポート有りのイメージを使用します。
[1]	こちらを参考に、NVIDIA Container Toolkit をインストールしておきます。
[2]	TensorFlow Docker (GPU) の root ユーザーでの利用例です。当例では CUDA 10.1, TensorFlow 2.1.0 を使用します。一般ユーザーで [docker] コマンドを使用したい場合は、該当ユーザーを [docker] グループに所属させる必要があります。 TensorFlow 公式のテスト済みのビルド構成は以下から確認可能です。 ⇒ https://www.tensorflow.org/install/source?hl=en#gpu_support_2

# CUDA/TensorFlow イメージを Pull

[root@dlp ~]#

docker pull nvidia/cuda:10.1-base

[root@dlp ~]#

docker pull tensorflow/tensorflow:2.1.0-gpu-py3

[root@dlp ~]#

docker images

REPOSITORY              TAG                 IMAGE ID            CREATED             SIZE
tensorflow/tensorflow   2.1.0-gpu-py3       e2a4af785bdb        6 months ago        4.11GB
nvidia/cuda             10.1-base           3b55548ae91f        7 months ago        106MB

# [nvidia-smi] 動作確認

[root@dlp ~]#

docker run --gpus all --rm nvidia/cuda:10.1-base nvidia-smi

Fri Jul 24 02:31:04 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 450.57       Driver Version: 450.57       CUDA Version: 11.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce GTX 1070    Off  | 00000000:05:00.0 Off |                  N/A |
| 27%   38C    P5    24W / 180W |      0MiB /  8119MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

# TensorFlow 動作確認

[root@dlp ~]#

docker run --gpus all --rm tensorflow/tensorflow:2.1.0-gpu-py3 \
python -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"

2020-07-24 02:31:47.674743: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libnvinfer.so.6
2020-07-24 02:31:47.677653: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libnvinfer_plugin.so.6
2020-07-24 02:31:48.397372: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcuda.so.1
2020-07-24 02:31:48.627649: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-24 02:31:48.628521: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1555] Found device 0 with properties:
pciBusID: 0000:05:00.0 name: GeForce GTX 1070 computeCapability: 6.1
coreClock: 1.7845GHz coreCount: 15 deviceMemorySize: 7.93GiB deviceMemoryBandwidth: 238.66GiB/s
2020-07-24 02:31:48.628584: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
2020-07-24 02:31:48.628642: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10
2020-07-24 02:31:48.633279: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10
2020-07-24 02:31:48.633882: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10
2020-07-24 02:31:48.638866: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolver.so.10
2020-07-24 02:31:48.640898: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10
2020-07-24 02:31:48.640963: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7
2020-07-24 02:31:48.641109: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-24 02:31:48.642019: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-24 02:31:48.642878: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1697] Adding visible gpu devices: 0
2020-07-24 02:31:48.655202: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 2801765000 Hz
2020-07-24 02:31:48.655964: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x54ced50 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-07-24 02:31:48.656014: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version
2020-07-24 02:31:48.726065: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-24 02:31:48.726918: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x5544a30 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2020-07-24 02:31:48.726963: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): GeForce GTX 1070, Compute Capability 6.1
2020-07-24 02:31:48.727437: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-24 02:31:48.728399: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1555] Found device 0 with properties:
pciBusID: 0000:05:00.0 name: GeForce GTX 1070 computeCapability: 6.1
coreClock: 1.7845GHz coreCount: 15 deviceMemorySize: 7.93GiB deviceMemoryBandwidth: 238.66GiB/s
2020-07-24 02:31:48.728487: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
2020-07-24 02:31:48.728528: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10
2020-07-24 02:31:48.728581: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10
2020-07-24 02:31:48.728625: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10
2020-07-24 02:31:48.728671: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolver.so.10
2020-07-24 02:31:48.728717: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10
2020-07-24 02:31:48.728755: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7
2020-07-24 02:31:48.728907: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-24 02:31:48.729950: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-24 02:31:48.730871: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1697] Adding visible gpu devices: 0
2020-07-24 02:31:48.730963: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
2020-07-24 02:31:48.983849: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1096] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-07-24 02:31:48.983900: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1102]      0
2020-07-24 02:31:48.983917: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] 0:   N
2020-07-24 02:31:48.984176: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-24 02:31:48.984850: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-24 02:31:48.985529: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1241] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 7588 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1070, pci bus id: 0000:05:00.0, compute capability: 6.1)
tf.Tensor(-2189.8442, shape=(), dtype=float32)