Don't have Telegram yet? Try it now!
https://hgpu.org/?p=8509
Use of CUDA for the Continuous Space Language Model