标签: gpu

我正在尝试安装CUDA,但是我收到一条消息"没有找到支持的visual studio版本".我认为这是因为我使用的是Visual Studio 2017(社区),而CUDA目前仅支持Visual Studio 2015.不幸的是,微软不允许我在不支付订阅费的情况下下载旧版本的Visual Studio.

有没有办法解决VS 2017的兼容性问题,还是我不能使用CUDA？

I am applying transfer-learning on a pre-trained network using the GPU version of keras. I don't understand how to define the parameters max_queue_size, workers, and use_multiprocessing. If I change these parameters (primarily to speed-up learning), I am unsure whether all data is still seen per epoch.

max_queue_size:

• maximum size of the internal training queue which is used to "precache" samples from the generator

• Question: Does this refer to how many batches are prepared on CPU? How …

在训练循环中，我将一批数据加载到 CPU 中，然后将其传输到 GPU：

import torch.utils as utils

train_loader = utils.data.DataLoader(train_dataset, batch_size=128, shuffle=True, num_workers=4, pin_memory=True)

for inputs, labels in train_loader:
    inputs, labels = inputs.to(device), labels.to(device)

这种加载数据的方式非常耗时。有什么方法可以直接将数据加载到 GPU 中而不需要传输步骤吗？

我有一个推力device_vector.我想将它转换为原始指针,以便我可以将它传递给内核.我怎么能这样做？

thrust::device_vector<int> dv(10);
//CAST TO RAW
kernel<<<bl,tpb>>>(pass raw)

我曾经偶然发现一个术语"非连贯"和"连贯"的记忆

与图形编程相关的技术论文.我一直在寻找一个简单而明确的解释,但是发现了这种类型的"硬核"论文.我很高兴收到外行人的样式答案,关于GPU体系结构实际上是什么连贯的内存以及它是如何实现的与其他(可能不相干)的内存类型进行比较.

更新:仍然发生在Tensorflow 1.7.0中

更新:我写了一个协同合作的笔记本再现谷歌的GPU硬件这个错误:https://drive.google.com/file/d/13V87kSTyyFVMM7NoJNk9QTsCYS7FRbyz/view?usp=sharing

更新:在tf.gather对此问题的第一次修订中错误地指责之后,我现在将其缩小为tf.reduce_sum与占位符一起形成:

tf.reduce_sum 为大型张量生成零(仅在GPU上),其形状取决于占位符.

在向占位符提供大整数时运行以下代码batch_size(在我的情况下> 700000):

import tensorflow as tf
import numpy as np

graph = tf.Graph()
with graph.as_default():
    batch_size = tf.placeholder(tf.int32,shape=[])
    ones_with_placeholder = tf.ones([batch_size,256,4])
    sum_out = tf.reduce_sum(ones_with_placeholder,axis=2)
    min_sum_out = tf.reduce_min(sum_out)

sess = tf.Session(graph=graph)

sum_result,min_sum_result = sess.run([sum_out,min_sum_out],feed_dict={batch_size: 1000000})
print("Min value in sum_out processed on host with numpy:", np.min(sum_result))
print("Min value in sum_out tensor processed in graph with tf:", min_sum_result)

显示以下错误结果:

Min value in sum_out processed on host with …

我已经训练了3个模型,现在正在运行代码,按顺序加载3个检查点中的每一个并使用它们运行预测.我正在使用GPU.

加载第一个模型时,它会预先分配整个GPU内存(我希望通过第一批数据处理).但是当它完成时它不会卸载内存.当加载第二个模型时,使用两个tf.reset_default_graph()并且with tf.Graph().as_default()GPU内存仍然完全从第一个模型消耗,然后第二个模型缺乏内存.

有没有办法解决这个问题,除了使用Python子进程或多处理来解决问题(我通过谷歌搜索找到的唯一解决方案)？

我正在使用Ubuntu 14.04 LTS运行AWS EC2 g2.2xlarge实例.我想在训练我的TensorFlow模型时观察GPU的利用率.我试图运行'nvidia-smi'时遇到错误.

ubuntu@ip-10-0-1-213:/etc/alternatives$ cd /usr/lib/nvidia-375/bin
ubuntu@ip-10-0-1-213:/usr/lib/nvidia-375/bin$ ls
nvidia-bug-report.sh     nvidia-debugdump     nvidia-xconfig
nvidia-cuda-mps-control  nvidia-persistenced
nvidia-cuda-mps-server   nvidia-smi
ubuntu@ip-10-0-1-213:/usr/lib/nvidia-375/bin$ ./nvidia-smi
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.


ubuntu@ip-10-0-1-213:/usr/lib/nvidia-375/bin$ dpkg -l | grep nvidia 
ii  nvidia-346                                            352.63-0ubuntu0.14.04.1                             amd64        Transitional package for nvidia-346
ii  nvidia-346-dev                                        346.46-0ubuntu1                                     amd64        NVIDIA binary Xorg driver development files
ii  nvidia-346-uvm                                        346.96-0ubuntu0.0.1                                 amd64        Transitional package for nvidia-346
ii  nvidia-352                                            375.26-0ubuntu1                                     amd64        Transitional package for …

我试图运行下面的代码,但报告错误:

NvvmSupportError:找不到libNVVM.执行conda install cudatoolkit:找不到库nvvm

我的开发环境是:Ubuntu 17.04,Spyder/Python3.5,我通过conda(numba和cudatoolkit)安装.Nvidia GPU(GTX 1070和GTX 1060).

import numpy as np
from timeit import default_timer as timer
from numba import vectorize

@vectorize(["float32(float32, float32)"], target='cuda')     
def VecADD(a,b):
    return a+b        

n = 32000000
a = np.ones (n, dtype=np.float32) 
b = np.ones (n, dtype=np.float32)     
c = np.zeros(n, dtype=np.float32) 

start = timer()
C = VecADD(a,b)
print (timer() - start)

有谁知道如何解决这个问题？

我可以列出gpu设备唱下面的tensorflow代码:

import tensorflow as tf
from tensorflow.python.client import device_lib
print device_lib.list_local_devices()

结果是:

[name: "/device:CPU:0"
 device_type: "CPU"
 memory_limit: 268435456
 locality {
 }
 incarnation: 17897160860519880862, name: "/device:XLA_GPU:0"
 device_type: "XLA_GPU"
 memory_limit: 17179869184
 locality {
 }
 incarnation: 9751861134541508701
 physical_device_desc: "device: XLA_GPU device", name: "/device:XLA_CPU:0"
 device_type: "XLA_CPU"
 memory_limit: 17179869184
 locality {
 }
 incarnation: 5368380567397471193
 physical_device_desc: "device: XLA_CPU device", name: "/device:GPU:0"
 device_type: "GPU"
 memory_limit: 21366299034
 locality {
   bus_id: 1
   links {
     link {
       device_id: 1
       type: "StreamExecutor"
       strength: 1
     }
   }
 }
 incarnation: 7110958745101815531
 physical_device_desc: "device: …