ik_llama_docker/glm-5.1-IQ3_KS.sh

40 lines
1.0 KiB
Bash
Executable File

#!/bin/bash
#grab container image name
MODEL_NAME="GLM-5.1-IQ3_KW"
CONT_NAME="ik-$MODEL_NAME"
MODEL="GLM-5.1-IQ3_KS-00001-of-00008.gguf"
MODEL_FOLDER="/sam4t/ENC/LLM/GLM-5.1"
PORT=10987
# Check if llama_env.sh exists, if not run create_new_image.sh
if [ ! -f "./env/ik_env.sh" ]; then
echo "./env/ik_env.sh not found, running create_new_IK_image.sh..."
./create_new_IK_image.sh
fi
source ./env/ik_env.sh
echo $IMAGE
docker container create --name $CONT_NAME --network llms -p $PORT:$PORT -v $MODEL_FOLDER:/model --user $(id -u):$(id -g) --gpus=all --restart on-failure:3 \
--health-cmd "curl -f http://$CONT_NAME:$PORT/v1/models || exit 1" \
--health-interval 5s \
--health-timeout 2s \
--health-start-period 60s \
$IMAGE \
/ik_llama.cpp/build/bin/llama-server \
--model /model/$MODEL \
--alias $MODEL_NAME \
--ctx-size 64000 \
-amb 512 \
-muge \
--merge-qkv \
-ngl 999 \
--n-cpu-moe 66 \
--parallel 1 --threads 32 \
--host $CONT_NAME \
--port $PORT --jinja --no-mmap
echo "Done!"
echo $CONT_NAME