From 9087c50efabcfe2f3e5553bb861af3d01fca6c92 Mon Sep 17 00:00:00 2001 From: Mikko Juola Date: Mon, 13 Mar 2023 12:41:57 -0700 Subject: [PATCH] Add notes about improving sampler to README.md --- README.md | 3 +++ 1 file changed, 3 insertions(+) diff --git a/README.md b/README.md index f711b49..80e1f9c 100644 --- a/README.md +++ b/README.md @@ -67,3 +67,6 @@ This is a hobby thing for me so don't expect updates or help. * I've heard there is some thing called Tensor Cores on nVidia GPUs. Not accessible with OpenCL. But might be accessible on Vulkan with a an extension. +* More sophisticated token sampling. I saw on Hackernews some comments how the + samplers are kinda garbage and you can get much better results with good + defaults and things like repetition penalty.