Increased start time, removed cache

qwen3.5-397B-A17B-IQ4_KSS.sh

@@ -19,7 +19,7 @@ docker container create --name $CONT_NAME --network llms -p $PORT:$PORT -v $MODE
 	--health-cmd "curl -f http://$CONT_NAME:$PORT/v1/models || exit 1" \
 	--health-interval 5s \
 	--health-timeout 5s \
-	--health-start-period 20s \
+	--health-start-period 50s \
 	$IMAGE \
 	/ik_llama.cpp/build/bin/llama-server \
 	--model /model/$MODEL \
@@ -28,7 +28,7 @@ docker container create --name $CONT_NAME --network llms -p $PORT:$PORT -v $MODE
 	-ngl 999 \
 	--n-cpu-moe 38 \
 	--parallel 2 --threads 32 \
-	-cram 32768 -cram-n-min 128 -crs .9 --ctx-checkpoints 512 \
+	--cache-ram 0 \
 	--host $CONT_NAME \
 	--port $PORT --jinja --no-mmap