matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-17 11:17:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix clearing kv cache between requests with the same prompt (#2186)

Browse Source 
* Fix clearing kv cache between requests with the same prompt

* fix powershell script
This commit is contained in:
Jeffrey Morgan
2024-01-25 13:46:20 -08:00
committed by GitHub
parent 7c40a67841
commit a64570dcae
4 changed files with 65 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
llm/generate/gen_common.sh
View File

@@ -61,6 +61,17 @@ apply_patches() {
if ! grep ollama ${LLAMACPP_DIR}/examples/server/CMakeLists.txt; then
echo 'include (../../../ext_server/CMakeLists.txt) # ollama' >>${LLAMACPP_DIR}/examples/server/CMakeLists.txt
fi
# apply temporary patches until fix is upstream
for patch in ../patches/*.diff; do
for file in $(grep "^+++ " ${patch} | cut -f2 -d' ' | cut -f2- -d/); do
(cd ${LLAMACPP_DIR}; git checkout ${file})
done
done
for patch in ../patches/*.diff; do
(cd ${LLAMACPP_DIR} && git apply ${patch})
done
# Avoid duplicate main symbols when we link into the cgo binary
sed -e 's/int main(/int __main(/g' <${LLAMACPP_DIR}/examples/server/server.cpp >${LLAMACPP_DIR}/examples/server/server.cpp.tmp &&
mv ${LLAMACPP_DIR}/examples/server/server.cpp.tmp ${LLAMACPP_DIR}/examples/server/server.cpp