fix t_llm_request is unbounded problem (#657)

This commit is contained in:
Shuchang Zheng
2024-07-29 13:54:42 -07:00
committed by GitHub
parent 112269ce1d
commit dd188b5f4d

View File

@@ -190,12 +190,12 @@ class LLMAPIHandlerFactory:
}
).encode("utf-8"),
)
t_llm_request = time.perf_counter()
try:
# TODO (kerem): add a timeout to this call
# TODO (kerem): add a retry mechanism to this call (acompletion_with_retries)
# TODO (kerem): use litellm fallbacks? https://litellm.vercel.app/docs/tutorials/fallbacks#how-does-completion_with_fallbacks-work
LOG.info("Calling LLM API", llm_key=llm_key, model=llm_config.model_name)
t_llm_request = time.perf_counter()
response = await litellm.acompletion(
model=llm_config.model_name,
messages=messages,