#!/bin/bash
#grab container image name
MODEL_NAME="Qwen3.5-397B-A17B"
CONT_NAME="ik-$MODEL_NAME"

MODEL="huggingface/hub/models--ubergarm--Qwen3.5-397B-A17B-GGUF/snapshots/bb9c5e48cdcad8716ef0dfff1f48f43b34b89cb2/IQ4_KSS/Qwen3.5-397B-A17B-IQ4_KSS-00001-of-00006.gguf"
MODEL_FOLDER="/home/jared/.cache"

source ./ik_env.sh
echo $IK_IMAGE
docker container create --name $CONT_NAME --network llms -p 10987:10987 -v $MODEL_FOLDER:/model --user $(id -u):$(id -g) --gpus=all --restart on-failure:3 \
	--health-cmd "curl -f http://$CONT_NAME:10987/v1/models || exit 1" \
	--health-interval 5s \
	--health-timeout 5s \
	--health-start-period 20s \
	$IK_IMAGE \
	/ik_llama.cpp/build/bin/llama-server \
	--model /model/$MODEL \
	--alias $MODEL_NAME \
	-ctk q8_0 -ctv q8_0 \
	-ngl 999 \
	--n-cpu-moe 38 \
	--parallel 2 --threads 32 \
	--host $CONT_NAME \
	--port 10987 --jinja --no-mmap