Merge pull request #47 from sambarnes/sentry-generation

fix: report generation errors to sentry w/ model name

modal/runner/engines/vllm.py

@@ -1,3 +1,4 @@
+import logging
 from typing import Optional
 
 from modal import method
@@ -44,12 +45,12 @@ def __init__(self, params: VllmParams):
         from vllm.engine.arg_utils import AsyncEngineArgs
         from vllm.engine.async_llm_engine import AsyncLLMEngine
 
-        engine_args = AsyncEngineArgs(
+        self.engine_args = AsyncEngineArgs(
             **params.dict(),
             disable_log_requests=True,
         )
 
-        self.engine = AsyncLLMEngine.from_engine_args(engine_args)
+        self.engine = AsyncLLMEngine.from_engine_args(self.engine_args)
 
     # @method()
     # async def tokenize_prompt(self, payload: Payload) -> List[int]:
@@ -116,7 +117,9 @@ async def generate(self, payload: CompletionPayload, params):
             print(f"Request completed: {throughput:.4f} tokens/s")
         except Exception as err:
             e = create_error_text(err)
-            print(e)
+            logging.exception(
+                "Failed generation", extra={"model": self.engine_args.model}
+            )
             if payload.stream:
                 yield create_sse_data(e)
             else: