GPUs - BKJackson/BKJackson_Wiki GitHub Wiki
Scaling Up with GPUs
From bare metal to a 70B model: infrastructure set-up and scripts
GPU's can be programmed
- CUDA (NVDIA only)
- Write C code that runs directly on the GPU
- Higher-level APIs: cuBLAS, cuFFT, cuDNN, etc
- OpenCL
- Similar to CUDA, but runs on anything
- Usually slower :(
- Udacity: Intro to Parallel Programming http://www.udacity.com/cource/cs344
- For deep learning use existing libraries
Credit for above: Li, Karpathy, and Johnson (2016)