Using GPU

Prerequisites

  • At least one CUDA compatible GPU. For a list of compatible GPUs, see NVIDIA's guide.

  • NVIDIA developer account. If you don't have one, create a free account.

  • Make sure the appropriate driver is installed for the GPU.

Install dependencies

  1. Install CUDA v10.1. Make sure you install CUDA v10.1, not any other newer version.

  2. Install cuDNN v7.6.4 for CUDA 10.1. You cannot have multiple versions of cuDNN installed. After downloading cuDNN v7.6.4 zip file and unpacking it, copy <CUDNN_zip_files_path>\cuda\bin\cudnn64_7.dll to <YOUR_DRIVE>\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin.

Troubleshooting

How do I know what GPU I have?

Follow the instructions provided:

  1. Right-click on desktop

  2. If you see "NVIDIA Control Panel" or "NVIDIA Display" in the pop-up window, you have an NVIDIA GPU

  3. Click on "NVIDIA Control Panel" or "NVIDIA Display" in the pop-up window

  4. Look at "Graphics Card Information"

  5. You will see the name of your NVIDIA GPU

I don't see NVIDIA Control Panel (or it fails to open) but I know I have an NVIDIA GPU.

  1. Open Device Manager

  2. Look at Display adapters

  3. Install the appropriate driver for your GPU.

How do I see what version of CUDA I have?

  1. Open a PowerShell or command line window Type in nvcc --version

PreviousHow to train modelsNextModel Evaluation

Last updated