Making sure bool type is passed as bool in our tests

Author: oandreeva-nv

ci/L0_multi_gpu/multi_lora/test.sh

@@ -55,12 +55,12 @@ export SERVER_ENABLE_LORA=true
 model_json=$(cat <<EOF
 {
     "model":"./weights/backbone/gemma-2b",
-    "disable_log_requests": "true",
+    "disable_log_requests": true,
     "gpu_memory_utilization": 0.7,
     "tensor_parallel_size": 2,
     "block_size": 16,
-    "enforce_eager": "true",
-    "enable_lora": "true",
+    "enforce_eager": true,
+    "enable_lora": true,
     "max_lora_rank": 32,
     "lora_extra_vocab_size": 256,
     "distributed_executor_backend":"ray"
@@ -115,12 +115,12 @@ export SERVER_ENABLE_LORA=false
 model_json=$(cat <<EOF
 {
     "model":"./weights/backbone/gemma-2b",
-    "disable_log_requests": "true",
+    "disable_log_requests": true,
     "gpu_memory_utilization": 0.8,
     "tensor_parallel_size": 2,
     "block_size": 16,
-    "enforce_eager": "true",
-    "enable_lora": "false",
+    "enforce_eager": true,
+    "enable_lora": false,
     "lora_extra_vocab_size": 256,
     "distributed_executor_backend":"ray"
 }

samples/model_repository/vllm_model/config.pbtxt

@@ -35,3 +35,10 @@ instance_group [
     kind: KIND_MODEL
   }
 ]
+
+parameters: {
+  key: "REPORT_CUSTOM_METRICS"
+  value: {
+    string_value:"yes"
+  }
+}

src/model.py

@@ -187,9 +187,19 @@ def init_engine(self):
     def setup_lora(self):
         self.enable_lora = False
 
-        if (
-            "enable_lora" in self.vllm_engine_config.keys()
-            and self.vllm_engine_config["enable_lora"].lower() == "true"
+        # Check if `enable_lora` field is in the `model.json`,
+        # and if it is, read its contents, which can be string or bool.
+        if "enable_lora" in self.vllm_engine_config.keys() and (
+            (
+                isinstance(self.vllm_engine_config["enable_lora"], str)
+                and self.vllm_engine_config["enable_lora"].lower() == "true"
+            )
+            or (
+                (
+                    isinstance(self.vllm_engine_config["enable_lora"], bool)
+                    and self.vllm_engine_config["enable_lora"]
+                )
+            )
         ):
             # create Triton LoRA weights repository
             multi_lora_args_filepath = os.path.join(