ik_llama_docker/qwen3.5-397B-A17B-IQ2_KL-CPU.sh

37 lines
1.1 KiB
Bash
Executable File

#!/bin/bash
#grab container image name
MODEL_NAME="Qwen3.5-397B-A17B"
CONT_NAME="LLM-$MODEL_NAME"
MODEL="huggingface/hub/models--ubergarm--Qwen3.5-397B-A17B-GGUF/snapshots/bb9c5e48cdcad8716ef0dfff1f48f43b34b89cb2/IQ2_KL/Qwen3.5-397B-A17B-IQ2_KL-00001-of-00004.gguf"
MODEL_FOLDER="/home/jared/.cache"
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 5s \
--health-start-period 50s \
$IMAGE \
/ik_llama.cpp/build/bin/llama-server \
--model /model/$MODEL \
--alias $MODEL_NAME \
-ctk q8_0 -ctv q8_0 \
-ngl 999 \
--parallel 2 --threads 24 \
--host $CONT_NAME \
--port $PORT --jinja --no-mmap
echo "Done!"
echo $CONT_NAME
docker container start -a $CONT_NAME