From 37659d2c4e38adebf3811d0997c8a79e9719edff Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Thu, 1 Jun 2023 22:33:50 +0800
Subject: [PATCH] allow blasbatchsize -1 which disables blas, but keeps
 benefits like gpu offloads.

---
 gpttype_adapter.cpp | 2 +-
 koboldcpp.py        | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/gpttype_adapter.cpp b/gpttype_adapter.cpp
index 86c13ce25..483a1f957 100644
--- a/gpttype_adapter.cpp
+++ b/gpttype_adapter.cpp
@@ -791,7 +791,7 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
                             file_format == FileFormat::GPTJ_2 || 
                             file_format == FileFormat::RWKV_1 || 
                             file_format==FileFormat::RWKV_2);
-    bool blasmode = (approved_format && embd_inp.size() >= 32 && ggml_cpu_has_blas());
+    bool blasmode = (approved_format && embd_inp.size() >= 32 && ggml_cpu_has_blas() && blasbatchsize!=-1);
     // bool blasmode = false;
     int original_batch = params.n_batch;
     int original_threads = params.n_threads;
diff --git a/koboldcpp.py b/koboldcpp.py
index a34702bfc..f379b46a9 100644
--- a/koboldcpp.py
+++ b/koboldcpp.py
@@ -210,7 +210,7 @@ maxctx = 2048
 maxlen = 256
 modelbusy = False
 defaultport = 5001
-KcppVersion = "1.27"
+KcppVersion = "1.28"
 
 class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
     sys_version = ""
@@ -648,7 +648,7 @@ if __name__ == '__main__':
     parser.add_argument("--psutil_set_threads", help="Experimental flag. If set, uses psutils to determine thread count based on physical cores.", action='store_true')
     parser.add_argument("--highpriority", help="Experimental flag. If set, increases the process CPU priority, potentially speeding up generation. Use caution.", action='store_true')
     parser.add_argument("--contextsize", help="Controls the memory allocated for maximum context size, only change if you need more RAM for big contexts. (default 2048)", type=int,choices=[512,1024,2048,4096,8192], default=2048)
-    parser.add_argument("--blasbatchsize", help="Sets the batch size used in BLAS processing (default 512)", type=int,choices=[32,64,128,256,512,1024], default=512)
+    parser.add_argument("--blasbatchsize", help="Sets the batch size used in BLAS processing (default 512). Setting it to -1 disables BLAS mode, but keeps other benefits like GPU offload.", type=int,choices=[-1,32,64,128,256,512,1024], default=512)
     parser.add_argument("--stream", help="Uses pseudo streaming when generating tokens. Only for the Kobold Lite UI.", action='store_true')
     parser.add_argument("--smartcontext", help="Reserving a portion of context to try processing less frequently.", action='store_true')
     parser.add_argument("--unbantokens", help="Normally, KoboldAI prevents certain tokens such as EOS and Square Brackets. This flag unbans them.", action='store_true')