server: remove warmup methods everywhere

server/reflector/llm/base.py

@@ -1,17 +1,15 @@
 import importlib
 import json
 import re
-from time import monotonic
 from typing import TypeVar
 
 import nltk
 from prometheus_client import Counter, Histogram
-from transformers import GenerationConfig
-
 from reflector.llm.llm_params import TaskParams
 from reflector.logger import logger as reflector_logger
 from reflector.settings import settings
 from reflector.utils.retry import retry
+from transformers import GenerationConfig
 
 T = TypeVar("T", bound="LLM")
 
@@ -112,20 +110,6 @@ class LLM:
         self.m_generate_success = self.m_generate_success.labels(name)
         self.m_generate_failure = self.m_generate_failure.labels(name)
 
-    async def warmup(self, logger: reflector_logger):
-        start = monotonic()
-        name = self.__class__.__name__
-        logger.info(f"LLM[{name}] warming up...")
-        try:
-            await self._warmup(logger=logger)
-            duration = monotonic() - start
-            logger.info(f"LLM[{name}] warmup took {duration:.2f} seconds")
-        except Exception:
-            logger.exception(f"LLM[{name}] warmup failed, ignoring")
-
-    async def _warmup(self, logger: reflector_logger):
-        pass
-
     @property
     def tokenizer(self):
         """

server/reflector/llm/llm_modal.py

@@ -1,10 +1,9 @@
 import httpx
-from transformers import AutoTokenizer, GenerationConfig
-
 from reflector.llm.base import LLM
 from reflector.logger import logger as reflector_logger
 from reflector.settings import settings
 from reflector.utils.retry import retry
+from transformers import AutoTokenizer, GenerationConfig
 
 
 class ModalLLM(LLM):
@@ -12,7 +11,6 @@ class ModalLLM(LLM):
         super().__init__()
         self.timeout = settings.LLM_TIMEOUT
         self.llm_url = settings.LLM_URL + "/llm"
-        self.llm_warmup_url = settings.LLM_URL + "/warmup"
         self.headers = {
             "Authorization": f"Bearer {settings.LLM_MODAL_API_KEY}",
         }
@@ -27,15 +25,6 @@ class ModalLLM(LLM):
         # Replace this with a HTTP call
         return ["lmsys/vicuna-13b-v1.5"]
 
-    async def _warmup(self, logger):
-        async with httpx.AsyncClient() as client:
-            response = await client.post(
-                self.llm_warmup_url,
-                headers=self.headers,
-                timeout=60 * 5,
-            )
-            response.raise_for_status()
-
     async def _generate(
         self, prompt: str, gen_schema: dict | None, gen_cfg: dict | None, **kwargs
     ):