prep to have a cpu and gpu version

.lagoon.yml

@@ -1,6 +1 @@
-docker-compose-yaml: docker-compose.lagoon.yml
-environments:
-  main:
-    routes:
-      - llama2-api:
-        - llama2.marco-test6.amazeeio.review
+docker-compose-yaml: docker-compose.gpu.yml

docker-compose.yml

@@ -4,6 +4,11 @@ services:
     build:
       context: .
       dockerfile: Dockerfile
+    labels:
+      lagoon.type: basic-persistent
+      lagoon.persistent: /data
+      lagoon.service.port: 8000
+      lagoon.autogeneratedroute: true
     volumes:
       - ./data:/data:delegated
     ports:
@@ -14,5 +19,7 @@ services:
     build:
       context: .
       dockerfile: ui.Dockerfile
+    labels:
+      lagoon.type: node
     ports:
       - "3000:3000"

start-llama2.sh

@@ -5,4 +5,4 @@ ulimit -l unlimited
 echo "Model: ${MODEL}"
 
 python3 hug_model.py -s ${MODEL} -f "q5_1"
-python3 -B -m llama_cpp.server --model /data/model.bin --n_gpu_layers=43
+python3 -B -m llama_cpp.server --model /data/model.bin ${LLAMA_CPP_ARGS}