Tested on a Raspberry Pi 4 with 8GB RAM.
Whatever else you do, make sure you are using a 64-bit OS. The combination of the arm archtecture and a 32-bit OS will greatly limit package support.
Rasberry Pi OS is based on Debian, and very recently released a 64-bit version.
Ubuntu for Raspberry Pi 64-bit is another choice.
| conda create --name textgenrnn python | |
| conda activate textgenrnn | |
| conda install cudatoolkit==10.2.89 | |
| # this special conda source is vital to get the correct 8.1.x version of cudnn: | |
| conda install -c conda-forge cudnn | |
| git clone https://github.com/minimaxir/textgenrnn.git | |
| cd textgenrnn | |
| pip install -r requirements.txt |