From 0cca0726fef36732d640a95ce140da67325c7ce1 Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Sun, 23 Jul 2023 09:59:34 +0800
Subject: [PATCH] reduce number of retries, fixed maxlength > maxctx bug

---
 koboldcpp.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/koboldcpp.py b/koboldcpp.py
index 8143960ff..d704bc653 100755
--- a/koboldcpp.py
+++ b/koboldcpp.py
@@ -222,6 +222,8 @@ def generate(prompt,max_length=20, max_context_length=512, temperature=0.8, top_
     inputs = generation_inputs()
     outputs = ctypes.create_unicode_buffer(ctypes.sizeof(generation_outputs))
     inputs.prompt = prompt.encode("UTF-8")
+    if max_length >= max_context_length:
+        max_length = max_context_length-1
     inputs.max_context_length = max_context_length   # this will resize the context buffer if changed
     inputs.max_length = max_length
     inputs.temperature = temperature
@@ -1426,7 +1428,7 @@ def run_horde_worker(args, api_key, worker_name):
                     break
                 else:
                     currentjob_attempts += 1
-                    if currentjob_attempts>10:
+                    if currentjob_attempts>5:
                         break
             print("Server Busy - Not ready to generate...")
             time.sleep(5)