From a5cfeb7763e7c2c5833fde1095ed2e64466cb6d8 Mon Sep 17 00:00:00 2001
From: Andrei Betlen <abetlen@gmail.com>
Date: Thu, 15 Feb 2024 15:17:30 -0500
Subject: [PATCH] feat: Update llama.cpp

---
 llama_cpp/llama_cpp.py | 9 +++++++++
 vendor/llama.cpp       | 2 +-
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/llama_cpp/llama_cpp.py b/llama_cpp/llama_cpp.py
index 9979a67..13daadb 100644
--- a/llama_cpp/llama_cpp.py
+++ b/llama_cpp/llama_cpp.py
@@ -230,6 +230,15 @@ LLAMA_ROPE_SCALING_LINEAR = 1
 LLAMA_ROPE_SCALING_YARN = 2
 LLAMA_ROPE_SCALING_MAX_VALUE = LLAMA_ROPE_SCALING_YARN
 
+# enum llama_pooling_type {
+#     LLAMA_POOLING_NONE = 0,
+#     LLAMA_POOLING_MEAN = 1,
+#     LLAMA_POOLING_CLS  = 2,
+# };
+LLAMA_POOLING_NONE = 0
+LLAMA_POOLING_MEAN = 1
+LLAMA_POOLING_CLS = 2
+
 # enum llama_split_mode {
 #     LLAMA_SPLIT_NONE    = 0, // single GPU
 #     LLAMA_SPLIT_LAYER   = 1, // split layers and KV across GPUs
diff --git a/vendor/llama.cpp b/vendor/llama.cpp
index 8084d55..4524290 160000
--- a/vendor/llama.cpp
+++ b/vendor/llama.cpp
@@ -1 +1 @@
-Subproject commit 8084d554406b767d36b3250b3b787462d5dd626f
+Subproject commit 4524290e87b8e107cc2b56e1251751546f4b9051