Increase model wait time for vLLM

PyWorker Error Handling

lib/backend.py

@@ -30,7 +30,7 @@ from lib.data_types import (
     BenchmarkResult
 )
 
-VERSION = "0.2.0"
+VERSION = "0.2.1"
 
 MSG_HISTORY_LEN = 100
 log = logging.getLogger(__file__)

workers/openai/server.py

@@ -11,6 +11,7 @@ MODEL_SERVER_START_LOG_MSG = [
     "llama runner started",  # Ollama
     '"message":"Connected","target":"text_generation_router"',  # TGI
     '"message":"Connected","target":"text_generation_router::server"',  # TGI
+    "main: model loaded" # llama.cpp
 ]
 
 MODEL_SERVER_ERROR_LOG_MSGS = [
@@ -34,6 +35,7 @@ backend = Backend(
     model_server_url=os.environ["MODEL_SERVER_URL"],
     model_log_file=os.environ["MODEL_LOG"],
     allow_parallel_requests=True,
+    max_wait_time=600.0,
     benchmark_handler=CompletionsHandler(benchmark_runs=3, benchmark_words=256),
     log_actions=[
         *[(LogAction.ModelLoaded, info_msg) for info_msg in MODEL_SERVER_START_LOG_MSG],