README

Support for CUDA is ALWAYS on the way.

Installation

1. Download .whl files form Releases

2. Install

   pip install dlib-XX.XX.X-cpXX-cpXX-win_amd64.whl

   The XX depends on your situation.

---

Configuration

Open file C:\Users\XXX\AppData\Local\Programs\Python\PythonXX\Lib\site-packages\dlib\_init_.py.

The XXX depends on your situation and the whole filepath may be different based on your installation configuration.

With CUDA

If you use CUDA, configure the code:

if "ON" == "ON":
    add_lib_to_dll_path(
        "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/vXX.X/lib/x64/cudnn.lib"
    )
    add_lib_to_dll_path(
        "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/vXX.X/lib/x64/cudart.lib"
    )

The XX.X depends on your situation or the whole filepath may be different based on your installation configuration.

Without CUDA

If you do not use CUDA, delete the code:

if "ON" == "ON":
    add_lib_to_dll_path(
        "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/vXX.X/lib/x64/cudnn.lib"
    )
    add_lib_to_dll_path(
        "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/vXX.X/lib/x64/cudart.lib"
    )

Probably. I have not tried it.

---

Verification

import dlib

print(dlib.DLIB_USE_CUDA)
print(dlib.cuda.get_num_devices())

Or

Run the tempCodeRunnerFile.py.

---

Warning

Whether the .whl file works on your computer is a matter of luck!

If the .whl file smooths your installation, you could send an email of gratitude to me. This is enough to delight me for not less than a day and makes you incredibly sexually attractive.

If it does not work, please make an attempt at the following steps.

1. Try to install different versions of .whl files on the Release page.

2. Try .whl files on Dlib_Windows_Python3.x.

3. Download the source code from dlib (official) and install it by yourself.

By the way, you could also send an email to me to complain.

Note

My email address is shown on my GitHub profile.

PS

Configuration for CUDNN Installation

The installation of cuDNN greater than 9.0.0 may cause weird errors. Please check out the issue davisking/dlib#2979.

The following steps may work:

1. Move files from subfolders to parent folders.

   • Move all files from C:\Program Files\NVIDIA\CUDNN\vx.x\bin\xx.x to C:\Program Files\NVIDIA\CUDNN\vx.x\bin
   • Move all files from C:\Program Files\NVIDIA\CUDNN\vx.x\include\xx.x to C:\Program Files\NVIDIA\CUDNN\vx.x\inlcude
   • Move all files from C:\Program Files\NVIDIA\CUDNN\vx.x\lib\xx.x to C:\Program Files\NVIDIA\CUDNN\vx.x\lib

2. Add the system variable CMAKE_PRIFIX_PATH whose value is C:/Program Files/NVIDIA/CUDNN/vx.x.