Updated to latest DJL LMI image

Author: dgallitelli

workshops/diy-agents-with-sagemaker-and-bedrock/0-setup/2-setup-sagemaker-endpoint.ipynb

@@ -44,18 +44,21 @@
     "endpoint_name = name_from_base(model_id.split(\"/\")[1]+\"-ep\")\n",
     "model = DJLModel(\n",
     "    name=model_name, role=role,\n",
-    "    image_uri=f\"763104351884.dkr.ecr.{boto3.Session().region_name}.amazonaws.com/djl-inference:0.33.0-lmi15.0.0-cu128-v1.3\",\n",
+    "    image_uri=f\"763104351884.dkr.ecr.{boto3.Session().region_name}.amazonaws.com/djl-inference:0.34.0-lmi16.0.0-cu128-v1.2\",\n",
     "    env={\n",
-    "        \"HF_MODEL_ID\": model_id, # config: https://qwen.readthedocs.io/en/latest/framework/function_call.html#vllm \n",
-    "        \"OPTION_MAX_MODEL_LEN\": f\"{1024*20}\",\n",
+    "        \"HF_MODEL_ID\": model_id,
+    "        \"OPTION_MAX_MODEL_LEN\": f\"{1024*16}\",\n",
     "        \"OPTION_QUANTIZE\": \"fp8\",\n",
-    "        # vllm serve {model_id} --enable-auto-tool-choice --tool-call-parser hermes\n",
     "        \"OPTION_ROLLING_BATCH\": \"vllm\",\n",
+    "        \"OPTION_DTYPE\": \"bf16\",\n",
+    "        \"SERVING_FAIL_FAST\": \"true",\n",
+    "        \"OPTION_ROLLING_BATCH\": \"disable",\n",
+    "        \"OPTION_ASYNC_MODE\": \"true",\n",
+    "        \"OPTION_ENTRYPOINT\": \"djl_python.lmi_vllm.vllm_async_service",\n",
     "        \"OPTION_ENABLE_AUTO_TOOL_CHOICE\": \"true\",\n",
     "        \"OPTION_TOOL_CALL_PARSER\": \"hermes\",\n",
-    "        # --enable-reasoning --reasoning-parser deepseek_r1\n",
-    "        # \"OPTION_ENABLE_REASONING\": \"true\",\n",
-    "        # \"OPTION_REASONING_PARSER\": \"qwen3\", # currently not available in djl lmi15\n",
+    "        \"OPTION_ENABLE_REASONING\": \"true\",\n",
+    "        \"OPTION_REASONING_PARSER\": \"qwen3\",\n",
     "    }\n",
     ")\n",
     "model.deploy(\n",