首页 > 解决方案 > Tensorflow GPU:无法创建 cuSolverDN 实例

问题描述

TensorFlow 14 + cuda 8.0 + cudnn6.0 + gcc4.8,训练进度:

2018-08-01 15:48:53.786570: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1045] 设备对等矩阵 2018-08-01 15:48:53.786874: I tensorflow/core/common_runtime/gpu /gpu_device.cc:1051] DMA: 0 1 2 3 2018-08-01 15:48:53.786881: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1061] 0: YYYY 2018-08-01 15:48 :53.786884: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1061] 1: YYYY 2018-08-01 15:48:53.786887: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1061] 2: YYYY 2018-08-01 15:48:53.786890: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1061] 3: YYYY 2018-08-01 15:48:53.786897: I tensorflow/core/common_runtime/gpu/gpu_device .cc:1120] 创建 TensorFlow 设备 (/device:GPU:0) -> (设备:0,名称:GeForce GTX 1080 Ti,pci 总线 ID:0000:17:00.0,计算能力:6.1)2018-08-01 15:48:53.786901:I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] 创建 TensorFlow 设备 (/device:GPU:1) -> (device: 1, name: GeForce GTX 1080 Ti, pci bus id: 0000:18:00.0,计算能力:6.1) 2018-08-01 15:48:53.786905: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] 创建 TensorFlow 设备 (/device:GPU:2) -> (device: 2, name : GeForce GTX 1080 Ti, pci bus id: 0000:65:00.0, 计算能力: 6.1) 2018-08-01 15:48:53.786908: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] 创建 TensorFlow 设备(/device:GPU:3) -> (device: 3, name: GeForce GTX 1080 Ti, pci bus id: 0000:b3:00.0, 计算能力: 6.1)786905:I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] 创建 TensorFlow 设备(/device:GPU:2)->(设备:2,名称:GeForce GTX 1080 Ti,pci 总线 ID:0000:65: 00.0,计算能力:6.1) 2018-08-01 15:48:53.786908: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] 创建 TensorFlow 设备 (/device:GPU:3) -> (device: 3 ,名称:GeForce GTX 1080 Ti,pci 总线 ID:0000:b3:00.0,计算能力:6.1)786905:I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] 创建 TensorFlow 设备(/device:GPU:2)->(设备:2,名称:GeForce GTX 1080 Ti,pci 总线 ID:0000:65: 00.0,计算能力:6.1) 2018-08-01 15:48:53.786908: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] 创建 TensorFlow 设备 (/device:GPU:3) -> (device: 3 ,名称:GeForce GTX 1080 Ti,pci 总线 ID:0000:b3:00.0,计算能力:6.1)

可训练参数数量:40564970

num_total_steps: 543300

2018-08-01 15:51:36.092723: I tensorflow/core/kernels/cuda_solvers.cc:159] 为流 0xcd4b5330 创建 CudaSolver 句柄 2018-08-01 15:51:36.144979: F tensorflow/core/kernels/cuda_solvers。 cc:94] 检查失败:cusolverDnCreate(&cusolver_dn_handle) == CUSOLVER_STATUS_SUCCESS 未能创建 cuSolverDN 实例

我找不到有这个问题的人。但它减少了我的训练进度。拜托,谁能给我一些建议?

标签: pythontensorflowgpu

解决方案


推荐阅读