Update celery forwarder to use greenlets instead of processes

.ruff.toml

@@ -3,3 +3,6 @@ line-length = 100
 
 ignore = ["E501"]
 exclude = ["gen", "alembic"]
+
+[lint.per-file-ignores]
+"model-engine/model_engine_server/inference/forwarding/celery_forwarder.py" = ["E402"]

model-engine/model_engine_server/inference/forwarding/celery_forwarder.py

@@ -1,3 +1,7 @@
+from gevent import monkey
+
+monkey.patch_all()
+
 import argparse
 import json
 from datetime import datetime, timedelta
@@ -177,12 +181,7 @@ def start_celery_service(
         concurrency=concurrency,
         loglevel="INFO",
         optimization="fair",
-        # Don't use pool="solo" so we can send multiple concurrent requests over
-        # Historically, pool="solo" argument fixes the known issues of celery and some of the libraries.
-        # Particularly asyncio and torchvision transformers. This isn't relevant since celery-forwarder
-        # is quite lightweight
-        # TODO: we should probably use eventlet or gevent for the pool, since
-        # the forwarder is nearly the most extreme example of IO bound.
+        pool="gevent",
     )
     worker.start()
 

model-engine/requirements.in

@@ -25,6 +25,7 @@ ddtrace==1.8.3
 deprecation~=2.1
 docker~=5.0
 fastapi~=0.110.0
+gevent~=24.11.1
 gitdb2~=2.0
 gunicorn~=20.0
 httptools==0.5.0

model-engine/requirements.txt

@@ -167,6 +167,8 @@ frozenlist==1.3.3
     #   aiosignal
 fsspec==2023.10.0
     # via huggingface-hub
+gevent==24.11.1
+    # via -r model-engine/requirements.in
 gitdb==4.0.10
     # via gitpython
 gitdb2==2.0.6
@@ -175,8 +177,10 @@ gitpython==3.1.41
     # via -r model-engine/requirements.in
 google-auth==2.21.0
     # via kubernetes
-greenlet==2.0.2
-    # via sqlalchemy
+greenlet==3.1.1
+    # via
+    #   gevent
+    #   sqlalchemy
 gunicorn==20.1.0
     # via -r model-engine/requirements.in
 h11==0.14.0
@@ -569,10 +573,16 @@ yarl==1.9.2
     #   aiohttp
 zipp==3.16.0
     # via importlib-metadata
+zope-event==5.0
+    # via gevent
+zope-interface==7.2
+    # via gevent
 
 # The following packages are considered to be unsafe in a requirements file:
 setuptools==69.0.3
     # via
     #   gunicorn
     #   kubernetes
     #   kubernetes-asyncio
+    #   zope-event
+    #   zope-interface