ngxson HF staff commited on
Commit
04bf0e1
1 Parent(s): ce40819

Update start.sh

Browse files
Files changed (1) hide show
  1. start.sh +6 -2
start.sh CHANGED
@@ -5,13 +5,17 @@ if [ ! -d "llama.cpp" ]; then
5
  git clone https://github.com/ggerganov/llama.cpp
6
  fi
7
 
 
8
  if [[ -z "${RUN_LOCALLY}" ]]; then
9
  # enable CUDA if NOT running locally
10
- export LLAMA_CUDA=1
11
  fi
12
 
13
  cd llama.cpp
14
- make -j llama-quantize llama-gguf-split llama-imatrix
 
 
 
15
 
16
  cd ..
17
  python app.py
 
5
  git clone https://github.com/ggerganov/llama.cpp
6
  fi
7
 
8
+ export GGML_CUDA=OFF
9
  if [[ -z "${RUN_LOCALLY}" ]]; then
10
  # enable CUDA if NOT running locally
11
+ export GGML_CUDA=ON
12
  fi
13
 
14
  cd llama.cpp
15
+ cmake -B build -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=${GGML_CUDA} -DLLAMA_CURL=ON
16
+ cmake --build build --config Release -j --target llama-quantize llama-gguf-split llama-imatrix
17
+ cp ./build/bin/llama-* .
18
+ rm -rf build
19
 
20
  cd ..
21
  python app.py