diff --git a/README.md b/README.md
index be94d47..318699c 100644
--- a/README.md
+++ b/README.md
@@ -84,7 +84,7 @@ make dev
 ```
 or
 ```bash
-make dev-python  # Turn off triton backend (for non-cuda env developers)
+make dev-triton # Turn on triton backend (for cuda env developers)
 ```
 
 ## TODOs
diff --git a/deployment/docker-compose.python.yml b/deployment/docker-compose.python.yml
deleted file mode 100644
index dd48a5e..0000000
--- a/deployment/docker-compose.python.yml
+++ /dev/null
@@ -1,31 +0,0 @@
-version: '3.3'
-
-services:
-  init:
-    image: tabbyml/tabby
-    container_name: init
-    user: root
-    volumes:
-      - ${DATA_VOLUME}
-      - ${HF_VOLUME}
-    command: chown -R 1000 /data
-
-  tabby:
-    image: tabbyml/tabby
-    container_name: tabby
-    environment:
-      MODEL_NAME: ${MODEL_NAME}
-    ports:
-      - "5000:5000"
-      - "8080:8080"
-      - "8501:8501"
-    volumes:
-      - ${DATA_VOLUME}
-      - ${HF_VOLUME}
-    healthcheck:
-      test: ["CMD", "curl", "-f", "http://localhost:5000"]
-      interval: 2s
-      timeout: 2s
-      start_period: 1200s
-    depends_on:
-      - init
diff --git a/deployment/docker-compose.triton.yml b/deployment/docker-compose.triton.yml
new file mode 100644
index 0000000..e52a79b
--- /dev/null
+++ b/deployment/docker-compose.triton.yml
@@ -0,0 +1,15 @@
+version: '3.3'
+
+services:
+  tabby:
+    shm_size: 1gb
+    environment:
+      MODEL_BACKEND: triton
+      MODEL_REPLICA: ${MODEL_REPLICA:-1}
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: [gpu]
diff --git a/deployment/docker-compose.yml b/deployment/docker-compose.yml
index 771e08a..dd48a5e 100644
--- a/deployment/docker-compose.yml
+++ b/deployment/docker-compose.yml
@@ -13,11 +13,8 @@ services:
   tabby:
     image: tabbyml/tabby
     container_name: tabby
-    shm_size: 1gb
     environment:
       MODEL_NAME: ${MODEL_NAME}
-      MODEL_BACKEND: triton
-      MODEL_REPLICA: ${MODEL_REPLICA:-1}
     ports:
       - "5000:5000"
       - "8080:8080"
@@ -30,12 +27,5 @@ services:
       interval: 2s
       timeout: 2s
       start_period: 1200s
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - driver: nvidia
-              count: all
-              capabilities: [gpu]
     depends_on:
       - init
diff --git a/deployment/skypilot/default.yml b/deployment/skypilot/default.yml
index e1bd987..6e1aa06 100644
--- a/deployment/skypilot/default.yml
+++ b/deployment/skypilot/default.yml
@@ -22,4 +22,4 @@ setup: |
 
 run: |
   cd tabby/deployment
-  sudo MODEL_REPLICA=${MODEL_REPLICA:-8} docker-compose up
+  sudo MODEL_REPLICA=${MODEL_REPLICA:-8} docker-compose -f docker-compose.yml -f docker-compose.triton.yml up
diff --git a/development/Makefile b/development/Makefile
index b88b065..770edf0 100644
--- a/development/Makefile
+++ b/development/Makefile
@@ -7,5 +7,5 @@ build:
 dev:
 	docker-compose -f ../deployment/docker-compose.yml -f docker-compose.dev.yml $(UP_FLAGS)
 
-dev-python:
-	docker-compose -f ../deployment/docker-compose.python.yml -f docker-compose.dev.yml $(UP_FLAGS)
+dev-triton:
+	docker-compose -f ../deployment/docker-compose.yml -f ../deployment/docker-compose.triton.yml -f docker-compose.dev.yml $(UP_FLAGS)