Comment Comment by ioedward Normally people split up the model across multiple GPUs, i.e. model/tensor parallelism. parent ioedward 3 years ago
