add back depencencies changes

Signed-off-by: Gene Su <e870252314@gmail.com>

python/ray/llm/_internal/serve/deployments/llm/multiplex/utils.py

@@ -1,7 +1,10 @@
 import json
 import subprocess
+import time
 from typing import Any, Dict, List, Optional, Tuple, Union
 
+from asyncache import cached
+from cachetools import TLRUCache
 from fastapi import HTTPException
 from filelock import FileLock
 
@@ -190,6 +193,14 @@ def _get_object_from_cloud(object_uri: str) -> Union[str, object]:
         return body_str
 
 
+@cached(
+    cache=TLRUCache(
+        maxsize=4096,
+        getsizeof=lambda x: 1,
+        ttu=_validate_model_ttu,
+        timer=time.monotonic,
+    )
+)
 async def get_object_from_cloud(object_uri: str) -> Union[str, object]:
     """Calls _get_object_from_cloud with caching.
 

python/requirements/llm/llm-requirements.txt

@@ -1,2 +1,4 @@
 # Keep this in sync with the definition in setup.py for ray[llm]
 vllm>=0.7.2
+asyncache==0.3.1
+jsonref==1.1.0

python/requirements_compiled_rayllm_py311.txt

@@ -139,6 +139,12 @@ astor==0.8.1 \
     # via
     #   -c python/requirements_compiled_rayllm_test_py311.txt
     #   depyf
+asyncache==0.3.1 \
+    --hash=sha256:9a1e60a75668e794657489bdea6540ee7e3259c483517b934670db7600bf5035 \
+    --hash=sha256:ef20a1024d265090dd1e0785c961cf98b9c32cc7d9478973dcf25ac1b80011f5
+    # via
+    #   -c python/requirements_compiled_rayllm_test_py311.txt
+    #   -r python/requirements/llm/llm-requirements.txt
 attrs==25.1.0 \
     --hash=sha256:1c97078a80c814273a76b2a298a932eb681c87415c11dee0a6921de7f1b02c3e \
     --hash=sha256:c75a69e28a550a7e93789579c22aa26b0f5b83b75dc4e08fe092980051e1090a
@@ -247,6 +253,7 @@ cachetools==5.3.2 \
     --hash=sha256:861f35a13a451f94e301ce2bec7cac63e881232ccce7ed67fab9b5df4d3beaa1
     # via
     #   -c python/requirements_compiled_rayllm_test_py311.txt
+    #   asyncache
     #   google-auth
 certifi==2023.11.17 \
     --hash=sha256:9b469f3a900bf28dc19b8cfbf8019bf47f7fdd1a65a1d4ffb98fc14166beb4d1 \
@@ -1058,6 +1065,10 @@ jiter==0.8.2 \
     # via
     #   -c python/requirements_compiled_rayllm_test_py311.txt
     #   openai
+jsonref==1.1.0 \
+    --hash=sha256:32fe8e1d85af0fdefbebce950af85590b22b60f9e95443176adbde4e1ecea552 \
+    --hash=sha256:590dc7773df6c21cbf948b5dac07a72a251db28b0238ceecce0a2abfa8ec30a9
+    # via -r python/requirements/llm/llm-requirements.txt
 jsonschema==4.23.0 \
     --hash=sha256:d71497fef26351a33265337fa77ffeb82423f3ea21283cd9467bb03999266bc4 \
     --hash=sha256:fbadb6f8b144a8f8cf9f0b89ba94501d143e50411a1278633f56a7acf7fd5566

python/setup.py

@@ -361,7 +361,12 @@ def get_packages(self):
     #
     # Keep this in sync with python/requirements/llm/llm-requirements.txt
     #
-    setup_spec.extras["llm"] = list(set(["vllm>=0.7.2"] + setup_spec.extras["data"]))
+    setup_spec.extras["llm"] = list(
+        set(
+            ["vllm>=0.7.2", "asyncache==0.3.1", "jsonref==1.1.0"]
+            + setup_spec.extras["data"]
+        )
+    )
 
 # These are the main dependencies for users of ray. This list
 # should be carefully curated. If you change it, please reflect