So to prevent another dozen of answers suggesting invalid in the context of this thread suggestion to !kill -9 -1, let’s close this thread: The answer is simple: As of this writing Google simply gives only 5% of GPU to some of us, whereas 100% to the others. Period. dec-2019 update: The problem still exists … Read more
Set the environment variable CUDA_DEVICE_ORDER as: export CUDA_DEVICE_ORDER=PCI_BUS_ID Then the GPU IDs will be ordered by pci bus IDs.
A kernel launch is asynchronous. This means it returns control to the CPU thread immediately after starting up the GPU process, before the kernel has finished executing. So what is the next thing in the CPU thread here? Application exit. At application exit, it’s ability to send output to the standard output is terminated by … Read more
As I said in a comment, the overhead of invoking GPU kernels, and copying data to and from GPU, is very high. For operations on models with very little parameters it is not worth of using GPU since frequency of CPU cores is much higher. If you compare matrix multiplication (this is what DL mostly … Read more
According to https://community.amd.com/thread/169965 extern “C” { __declspec(dllexport) int AmdPowerXpressRequestHighPerformance = 1; } This will select the high performance GPU as long as no profile exists that assigns the application to another GPU. Please make sure to use a 13.35 or newer driver. Older drivers do not support this.
I am answering my own question. Easiest way to do is use connect to Local Runtime (https://research.google.com/colaboratory/local-runtimes.html) then select hardware accelerator as GPU as shown in (https://medium.com/deep-learning-turkey/google-colab-free-gpu-tutorial-e113627b9f5d).
These functions should help: >>> import torch >>> torch.cuda.is_available() True >>> torch.cuda.device_count() 1 >>> torch.cuda.current_device() 0 >>> torch.cuda.device(0) <torch.cuda.device at 0x7efce0b03be0> >>> torch.cuda.get_device_name(0) ‘GeForce GTX 950M’ This tells us: CUDA is available and can be used by one device. Device 0 refers to the GPU GeForce GTX 950M, and it is currently chosen by PyTorch.
Since CUDA 4.0 was released, multi-GPU computations of the type you are asking about are relatively easy. Prior to that, you would have need to use a multi-threaded host application with one host thread per GPU and some sort of inter-thread communication system in order to use mutliple GPUs inside the same host application. Now … Read more
Before running your code, run this shell command to tell torch that there are no GPUs: export CUDA_VISIBLE_DEVICES=”” This will tell it to use only one GPU (the one with id 0) and so on: export CUDA_VISIBLE_DEVICES=”0″
I do not code under Linux but in Windows (you are running it in emulator anyway) you can use WinAPI to directly access canvas of any window or even desktop from 3th party App. Some GPU overlays could be problematic to catch (especially DirectX based) but I had no problems with mine GL/GLSL for now. … Read more