From 6659652c9fd1853dcb2d1882efc8f14b159d5d43 Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Mon, 7 Aug 2023 11:05:06 +0800
Subject: [PATCH] lower actual temp used when temp=0

---
 gpttype_adapter.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/gpttype_adapter.cpp b/gpttype_adapter.cpp
index 4ae51ea23..a2283269c 100644
--- a/gpttype_adapter.cpp
+++ b/gpttype_adapter.cpp
@@ -243,7 +243,7 @@ void sample_temperature(llama_token_data_array * candidates_p, float temp)
     if (temp <= 0)
     {
         // Imitate greedy sampling
-        temp = 0.01f; //cannot be zero else div0
+        temp = 0.00390625f; //cannot be zero else div0, this is 1/256
         llama_sample_temperature(nullptr, candidates_p, temp);
         llama_sample_top_k(nullptr, candidates_p, 1, 1); //only want first candidate
     }