diff options
| author | Georgi Gerganov <ggerganov@gmail.com> | 2024-06-04 21:23:20 +0300 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-06-04 21:23:20 +0300 |
| commit | 554c247caffed64465f372661f2826640cb10430 (patch) | |
| tree | ce211403d5746a5cde175247a806223e83e17375 /scripts/server-llm.sh | |
| parent | 0cd6bd3483fa66124b76a8a8ac794d9ee18c70c1 (diff) | |
ggml : remove OpenCL (#7735)
ggml-ci
Diffstat (limited to 'scripts/server-llm.sh')
| -rw-r--r-- | scripts/server-llm.sh | 11 |
1 files changed, 3 insertions, 8 deletions
diff --git a/scripts/server-llm.sh b/scripts/server-llm.sh index eb6ce458..b3715e20 100644 --- a/scripts/server-llm.sh +++ b/scripts/server-llm.sh @@ -3,7 +3,7 @@ # Helper script for deploying llama.cpp server with a single Bash command # # - Works on Linux and macOS -# - Supports: CPU, CUDA, Metal, OpenCL +# - Supports: CPU, CUDA, Metal # - Can run all GGUF models from HuggingFace # - Can serve requests in parallel # - Always builds latest llama.cpp from GitHub @@ -19,7 +19,7 @@ # --port: port number, default is 8888 # --repo: path to a repo containing GGUF model files # --wtype: weights type (f16, q8_0, q4_0, q4_1), default is user-input -# --backend: cpu, cuda, metal, opencl, depends on the OS +# --backend: cpu, cuda, metal, depends on the OS # --gpu-id: gpu id, default is 0 # --n-parallel: number of parallel requests, default is 8 # --n-kv: KV cache size, default is 4096 @@ -72,7 +72,7 @@ function print_usage { printf " --port: port number, default is 8888\n" printf " --repo: path to a repo containing GGUF model files\n" printf " --wtype: weights type (f16, q8_0, q4_0, q4_1), default is user-input\n" - printf " --backend: cpu, cuda, metal, opencl, depends on the OS\n" + printf " --backend: cpu, cuda, metal, depends on the OS\n" printf " --gpu-id: gpu id, default is 0\n" printf " --n-parallel: number of parallel requests, default is 8\n" printf " --n-kv: KV cache size, default is 4096\n" @@ -387,9 +387,6 @@ elif [[ "$backend" == "cpu" ]]; then elif [[ "$backend" == "metal" ]]; then printf "[+] Building with Metal backend\n" make -j server $log -elif [[ "$backend" == "opencl" ]]; then - printf "[+] Building with OpenCL backend\n" - LLAMA_CLBLAST=1 make -j server $log else printf "[-] Unknown backend: %s\n" "$backend" exit 1 @@ -407,8 +404,6 @@ elif [[ "$backend" == "cpu" ]]; then args="-ngl 0" elif [[ "$backend" == "metal" ]]; then args="-ngl 999" -elif [[ "$backend" == "opencl" ]]; then - args="-ngl 999" else printf "[-] Unknown backend: %s\n" "$backend" exit 1 |