TF使用GPU计算 ========= ### 1. 检测显卡是否支持 CUDA --------------------------------- ### 2. 相关参考 ---------------- TensorFlow之CNN和GPU http://suanfazu.com/t/tensorflow-cnn-gpu/13216 cnn下，GPU是CPU的10倍以上，也就是CPU跑5W 5张图要100s，而在GPU上只要9s多。TensorFlow目前只支持nvidia卡 windows下安装Tensorflow+GPU加速 http://www.jianshu.com/p/c245d46d43f0 ubuntu16.04下安装TensorFlow(GPU加速) http://blog.csdn.net/zhaoyu106/article/details/52793183 Linux教程2: http://blog.csdn.net/zhaoyu106/article/details/52861268 AWS安装好驱动的镜像： Windows Server 2012 with NVIDIA GRID GPU Driver https://aws.amazon.com/marketplace/pp/B00FYCCNJ0/ref=portal_asin_url Windows Server 2008 R2 with NVIDIA GRID GPU Driver https://aws.amazon.com/marketplace/pp/B00FYCBRE2/ref=portal_asin_url Amazon Linux AMI with NVIDIA GRID and TESLA GPU Driver https://aws.amazon.com/marketplace/pp/B00FYCDDTE/ref=portal_asin_url AWS官方镜像GPU安装指南：http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/accelerated-computing-instances.html ### 3. 示例环境 ------------- Python 3.5 （64位） + Tensorflow-GPU版本 （目前只有3.5版本的GPU编译好的包） ~~~ pip install tensorflow-gpu ~~~ 卸载此前的 Tensorflow 版本 ~~~ pip uninstall tensorflow ~~~ Win下完整的安装步骤：  ### 4. 相关检测 -----------------  ### 5. GPU云计算资源 ------------------------------- NV GPU云计算提供商： http://www.nvidia.cn/object/gpu-cloud-computing-services-cn.html AWS机型：   AWS竞价示例降低成本：   实际经验来讲，一般弗吉尼亚、俄勒冈比较便宜  俄勒冈地区报价： （单位均为 每小时 费用）  弗吉尼亚报价：  竞价历史记录，了解可用时间：  建议使用的 Win镜像 （带有Desktop的Win）  设定使用价格策略：  请求成功提示：  请求成功后进入 实例 即可进行远程连接操作  ### 5. 实际跑 MINST 图像识别 官方示例 ------------------------------------------- 版本低于某一计算能力，被TF框架忽略 ~~~ C:\Users\XUN-2\Desktop\TF-DEV\TF_models\tutorials\image\mnist>python convolution al.py I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stre am_executor\dso_loader.cc:135] successfully opened CUDA library cublas64_80.dll locally I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stre am_executor\dso_loader.cc:126] Couldn't open CUDA library cudnn64_5.dll I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stre am_executor\cuda\cuda_dnn.cc:3517] Unable to load cuDNN DSO I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stre am_executor\dso_loader.cc:135] successfully opened CUDA library cufft64_80.dll l ocally I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stre am_executor\dso_loader.cc:135] successfully opened CUDA library nvcuda.dll local ly I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stre am_executor\dso_loader.cc:135] successfully opened CUDA library curand64_80.dll locally Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes. Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes. Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes. Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes. Extracting data\train-images-idx3-ubyte.gz Extracting data\train-labels-idx1-ubyte.gz Extracting data\t10k-images-idx3-ubyte.gz Extracting data\t10k-labels-idx1-ubyte.gz E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "BestSplits" device_type: "CPU"') fo r unknown op: BestSplits E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "CountExtremelyRandomStats" device_t ype: "CPU"') for unknown op: CountExtremelyRandomStats E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "FinishedNodes" device_type: "CPU"') for unknown op: FinishedNodes E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "GrowTree" device_type: "CPU"') for unknown op: GrowTree E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "ReinterpretStringToFloat" device_ty pe: "CPU"') for unknown op: ReinterpretStringToFloat E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "SampleInputs" device_type: "CPU"') for unknown op: SampleInputs E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "ScatterAddNdim" device_type: "CPU"' ) for unknown op: ScatterAddNdim E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "TopNInsert" device_type: "CPU"') fo r unknown op: TopNInsert E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "TopNRemove" device_type: "CPU"') fo r unknown op: TopNRemove E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "TreePredictions" device_type: "CPU" ') for unknown op: TreePredictions E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \framework\op_kernel.cc:943] OpKernel ('op: "UpdateFertileSlots" device_type: "C PU"') for unknown op: UpdateFertileSlots I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \common_runtime\gpu\gpu_device.cc:885] Found device 0 with properties: name: GeForce 710M major: 2 minor: 1 memoryClockRate (GHz) 1.55 pciBusID 0000:01:00.0 Total memory: 1.00GiB Free memory: 969.91MiB I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \common_runtime\gpu\gpu_device.cc:906] DMA: 0 I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \common_runtime\gpu\gpu_device.cc:916] 0: Y I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core \common_runtime\gpu\gpu_device.cc:948] Ignoring visible gpu device (device: 0, name: GeForce 710M, pci bus id: 0000:01:00.0) with Cuda compute capability 2.1. The minimum required Cuda capability is 3.0. Initialized! Step 0 (epoch 0.00), 21.5 ms Minibatch loss: 8.334, learning rate: 0.010000 Minibatch error: 85.9% Validation error: 84.6% Step 100 (epoch 0.12), 303.1 ms Minibatch loss: 3.254, learning rate: 0.010000 Minibatch error: 6.2% Validation error: 7.8% Step 200 (epoch 0.23), 303.9 ms Minibatch loss: 3.373, learning rate: 0.010000 Minibatch error: 9.4% Validation error: 4.3% Step 300 (epoch 0.35), 304.5 ms Minibatch loss: 3.150, learning rate: 0.010000 Minibatch error: 3.1% Validation error: 3.1% Step 400 (epoch 0.47), 304.7 ms Minibatch loss: 3.192, learning rate: 0.010000 Minibatch error: 4.7% Validation error: 2.9% Step 500 (epoch 0.58), 314.1 ms Minibatch loss: 3.176, learning rate: 0.010000 Minibatch error: 4.7% Validation error: 2.4% Step 600 (epoch 0.70), 307.6 ms Minibatch loss: 3.115, learning rate: 0.010000 Minibatch error: 3.1% Validation error: 2.0% Step 700 (epoch 0.81), 319.4 ms Minibatch loss: 2.969, learning rate: 0.010000 Minibatch error: 3.1% Validation error: 2.1% Step 800 (epoch 0.93), 327.6 ms Minibatch loss: 3.072, learning rate: 0.010000 Minibatch error: 6.2% Validation error: 2.1% Step 900 (epoch 1.05), 323.4 ms Minibatch loss: 2.898, learning rate: 0.009500 Minibatch error: 0.0% Validation error: 1.5% Step 1000 (epoch 1.16), 307.3 ms Minibatch loss: 2.855, learning rate: 0.009500 Minibatch error: 0.0% Validation error: 1.9% Step 1100 (epoch 1.28), 291.2 ms Minibatch loss: 2.824, learning rate: 0.009500 Minibatch error: 0.0% Validation error: 1.5% ~~~