mlserver - npm Package Compare versions

Comparing version 1.6.2rc1 to 1.7.0rc1

mlserver/grpc/dataplane_pb2_grpc.py

 # Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
 """Client and server classes corresponding to protobuf-defined services."""
 import grpc
+import warnings
 
 from . import dataplane_pb2 as dataplane__pb2
 
+GRPC_GENERATED_VERSION = "1.67.1"
+GRPC_VERSION = grpc.__version__
+_version_not_supported = False
 
+try:
+    from grpc._utilities import first_version_is_lower
+
+    _version_not_supported = first_version_is_lower(
+        GRPC_VERSION, GRPC_GENERATED_VERSION
+    )
+except ImportError:
+    _version_not_supported = True
+
+if _version_not_supported:
+    raise RuntimeError(
+        f"The grpc package installed is at version {GRPC_VERSION},"
+        + f" but the generated code in dataplane_pb2_grpc.py depends on"
+        + f" grpcio>={GRPC_GENERATED_VERSION}."
+        + f" Please upgrade your grpc module to grpcio>={GRPC_GENERATED_VERSION}"
+        + f" or downgrade your generated code using grpcio-tools<={GRPC_VERSION}."
+    )
+
+
 class GRPCInferenceServiceStub(object):
@@ -24,2 +47,3 @@ """
             response_deserializer=dataplane__pb2.ServerLiveResponse.FromString,
+            _registered_method=True,
         )
@@ -30,2 +54,3 @@ self.ServerReady = channel.unary_unary(
             response_deserializer=dataplane__pb2.ServerReadyResponse.FromString,
+            _registered_method=True,
         )
@@ -36,2 +61,3 @@ self.ModelReady = channel.unary_unary(
             response_deserializer=dataplane__pb2.ModelReadyResponse.FromString,
+            _registered_method=True,
         )
@@ -42,2 +68,3 @@ self.ServerMetadata = channel.unary_unary(
             response_deserializer=dataplane__pb2.ServerMetadataResponse.FromString,
+            _registered_method=True,
         )
@@ -48,2 +75,3 @@ self.ModelMetadata = channel.unary_unary(
             response_deserializer=dataplane__pb2.ModelMetadataResponse.FromString,
+            _registered_method=True,
         )
@@ -54,2 +82,3 @@ self.ModelInfer = channel.unary_unary(
             response_deserializer=dataplane__pb2.ModelInferResponse.FromString,
+            _registered_method=True,
         )
@@ -60,2 +89,3 @@ self.ModelStreamInfer = channel.stream_stream(
             response_deserializer=dataplane__pb2.ModelInferResponse.FromString,
+            _registered_method=True,
         )
@@ -66,2 +96,3 @@ self.RepositoryIndex = channel.unary_unary(
             response_deserializer=dataplane__pb2.RepositoryIndexResponse.FromString,
+            _registered_method=True,
         )
@@ -72,2 +103,3 @@ self.RepositoryModelLoad = channel.unary_unary(
             response_deserializer=dataplane__pb2.RepositoryModelLoadResponse.FromString,
+            _registered_method=True,
         )
@@ -78,2 +110,3 @@ self.RepositoryModelUnload = channel.unary_unary(
             response_deserializer=dataplane__pb2.RepositoryModelUnloadResponse.FromString,
+            _registered_method=True,
         )
@@ -206,2 +239,5 @@
     server.add_generic_rpc_handlers((generic_handler,))
+    server.add_registered_method_handlers(
+        "inference.GRPCInferenceService", rpc_method_handlers
+    )
 
@@ -243,2 +279,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -273,2 +310,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -303,2 +341,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -333,2 +372,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -363,2 +403,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -393,2 +434,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -423,2 +465,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -453,2 +496,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -483,2 +527,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -513,2 +558,3 @@
             metadata,
+            _registered_method=True,
         )

mlserver/grpc/dataplane_pb2.py

 # -*- coding: utf-8 -*-
 # Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
 # source: dataplane.proto
-# Protobuf Python Version: 4.25.1
+# Protobuf Python Version: 5.27.2
 """Generated protocol buffer code."""
 from google.protobuf import descriptor as _descriptor
 from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
 from google.protobuf import symbol_database as _symbol_database
 from google.protobuf.internal import builder as _builder
 
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC, 5, 27, 2, "", "dataplane.proto"
+)
 # @@protoc_insertion_point(imports)
@@ -23,12 +28,16 @@
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, "dataplane_pb2", _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-    _globals["DESCRIPTOR"]._options = None
+if not _descriptor._USE_C_DESCRIPTORS:
+    _globals["DESCRIPTOR"]._loaded_options = None
     _globals["DESCRIPTOR"]._serialized_options = b"\n\007ex.grpc\242\002\003HSW"
-    _globals["_MODELMETADATARESPONSE_TENSORMETADATA_PARAMETERSENTRY"]._options = None
     _globals[
         "_MODELMETADATARESPONSE_TENSORMETADATA_PARAMETERSENTRY"
+    ]._loaded_options = None
+    _globals[
+        "_MODELMETADATARESPONSE_TENSORMETADATA_PARAMETERSENTRY"
     ]._serialized_options = b"8\001"
-    _globals["_MODELMETADATARESPONSE_PARAMETERSENTRY"]._options = None
+    _globals["_MODELMETADATARESPONSE_PARAMETERSENTRY"]._loaded_options = None
     _globals["_MODELMETADATARESPONSE_PARAMETERSENTRY"]._serialized_options = b"8\001"
-    _globals["_MODELINFERREQUEST_INFERINPUTTENSOR_PARAMETERSENTRY"]._options = None
+    _globals["_MODELINFERREQUEST_INFERINPUTTENSOR_PARAMETERSENTRY"]._loaded_options = (
+        None
+    )
     _globals[
@@ -39,19 +48,21 @@ "_MODELINFERREQUEST_INFERINPUTTENSOR_PARAMETERSENTRY"
         "_MODELINFERREQUEST_INFERREQUESTEDOUTPUTTENSOR_PARAMETERSENTRY"
-    ]._options = None
+    ]._loaded_options = None
     _globals[
         "_MODELINFERREQUEST_INFERREQUESTEDOUTPUTTENSOR_PARAMETERSENTRY"
     ]._serialized_options = b"8\001"
-    _globals["_MODELINFERREQUEST_PARAMETERSENTRY"]._options = None
+    _globals["_MODELINFERREQUEST_PARAMETERSENTRY"]._loaded_options = None
     _globals["_MODELINFERREQUEST_PARAMETERSENTRY"]._serialized_options = b"8\001"
-    _globals["_MODELINFERRESPONSE_INFEROUTPUTTENSOR_PARAMETERSENTRY"]._options = None
     _globals[
         "_MODELINFERRESPONSE_INFEROUTPUTTENSOR_PARAMETERSENTRY"
+    ]._loaded_options = None
+    _globals[
+        "_MODELINFERRESPONSE_INFEROUTPUTTENSOR_PARAMETERSENTRY"
     ]._serialized_options = b"8\001"
-    _globals["_MODELINFERRESPONSE_PARAMETERSENTRY"]._options = None
+    _globals["_MODELINFERRESPONSE_PARAMETERSENTRY"]._loaded_options = None
     _globals["_MODELINFERRESPONSE_PARAMETERSENTRY"]._serialized_options = b"8\001"
-    _globals["_REPOSITORYMODELLOADREQUEST_PARAMETERSENTRY"]._options = None
+    _globals["_REPOSITORYMODELLOADREQUEST_PARAMETERSENTRY"]._loaded_options = None
     _globals["_REPOSITORYMODELLOADREQUEST_PARAMETERSENTRY"]._serialized_options = (
         b"8\001"
     )
-    _globals["_REPOSITORYMODELUNLOADREQUEST_PARAMETERSENTRY"]._options = None
+    _globals["_REPOSITORYMODELUNLOADREQUEST_PARAMETERSENTRY"]._loaded_options = None
     _globals["_REPOSITORYMODELUNLOADREQUEST_PARAMETERSENTRY"]._serialized_options = (
@@ -58,0 +69,0 @@ b"8\001"

mlserver/grpc/model_repository_pb2_grpc.py

 # Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
 """Client and server classes corresponding to protobuf-defined services."""
 import grpc
+import warnings
 
 from . import model_repository_pb2 as model__repository__pb2
 
+GRPC_GENERATED_VERSION = "1.67.1"
+GRPC_VERSION = grpc.__version__
+_version_not_supported = False
 
+try:
+    from grpc._utilities import first_version_is_lower
+
+    _version_not_supported = first_version_is_lower(
+        GRPC_VERSION, GRPC_GENERATED_VERSION
+    )
+except ImportError:
+    _version_not_supported = True
+
+if _version_not_supported:
+    raise RuntimeError(
+        f"The grpc package installed is at version {GRPC_VERSION},"
+        + f" but the generated code in model_repository_pb2_grpc.py depends on"
+        + f" grpcio>={GRPC_GENERATED_VERSION}."
+        + f" Please upgrade your grpc module to grpcio>={GRPC_GENERATED_VERSION}"
+        + f" or downgrade your generated code using grpcio-tools<={GRPC_VERSION}."
+    )
+
+
 class ModelRepositoryServiceStub(object):
@@ -21,2 +44,3 @@ """Missing associated documentation comment in .proto file."""
             response_deserializer=model__repository__pb2.RepositoryIndexResponse.FromString,
+            _registered_method=True,
         )
@@ -27,2 +51,3 @@ self.RepositoryModelLoad = channel.unary_unary(
             response_deserializer=model__repository__pb2.RepositoryModelLoadResponse.FromString,
+            _registered_method=True,
         )
@@ -33,2 +58,3 @@ self.RepositoryModelUnload = channel.unary_unary(
             response_deserializer=model__repository__pb2.RepositoryModelUnloadResponse.FromString,
+            _registered_method=True,
         )
@@ -81,2 +107,5 @@
     server.add_generic_rpc_handlers((generic_handler,))
+    server.add_registered_method_handlers(
+        "inference.model_repository.ModelRepositoryService", rpc_method_handlers
+    )
 
@@ -115,2 +144,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -145,2 +175,3 @@
             metadata,
+            _registered_method=True,
         )
@@ -175,2 +206,3 @@
             metadata,
+            _registered_method=True,
         )

mlserver/grpc/model_repository_pb2.py

 # -*- coding: utf-8 -*-
 # Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
 # source: model_repository.proto
-# Protobuf Python Version: 4.25.1
+# Protobuf Python Version: 5.27.2
 """Generated protocol buffer code."""
 from google.protobuf import descriptor as _descriptor
 from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
 from google.protobuf import symbol_database as _symbol_database
 from google.protobuf.internal import builder as _builder
 
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC, 5, 27, 2, "", "model_repository.proto"
+)
 # @@protoc_insertion_point(imports)
@@ -23,4 +28,4 @@
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, "model_repository_pb2", _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-    DESCRIPTOR._options = None
+if not _descriptor._USE_C_DESCRIPTORS:
+    DESCRIPTOR._loaded_options = None
     _globals["_REPOSITORYINDEXREQUEST"]._serialized_start = 54
@@ -27,0 +32,0 @@ _globals["_REPOSITORYINDEXREQUEST"]._serialized_end = 118

mlserver/parallel/registry.py

@@ -41,2 +41,8 @@ import asyncio
 
+def _append_gid_environment_hash(
+    env_hash: str, inference_pool_gid: Optional[str] = None
+) -> str:
+    return f"{env_hash}-{inference_pool_gid}"
+
+
 class InferencePoolRegistry:
@@ -84,3 +90,4 @@ """
             pool = await self._get_or_create_with_existing_env(
-                model.settings.parameters.environment_path
+                model.settings.parameters.environment_path,
+                model.settings.parameters.inference_pool_gid,
             )
@@ -92,3 +99,5 @@ else:
     async def _get_or_create_with_existing_env(
-        self, environment_path: str
+        self,
+        environment_path: str,
+        inference_pool_gid: Optional[str],
     ) -> InferencePool:
@@ -104,4 +113,9 @@ """
         env_hash = await compute_hash_of_string(expanded_environment_path)
+
+        if inference_pool_gid is not None:
+            env_hash = _append_gid_environment_hash(env_hash, inference_pool_gid)
+
         if env_hash in self._pools:
             return self._pools[env_hash]
+
         env = Environment(
@@ -121,9 +135,25 @@ env_path=expanded_environment_path,
         Creates or returns the InferencePool for a model that uses a
-        tarball as python environment.
+        tarball as a Python environment.
         """
         env_tarball = _get_env_tarball(model)
+        inference_pool_gid = (
+            model.settings.parameters.inference_pool_gid
+            if model.settings.parameters
+            else None
+        )
+
         if not env_tarball:
-            return self._default_pool
+            return (
+                self._pools.setdefault(
+                    inference_pool_gid,
+                    InferencePool(self._settings, on_worker_stop=self._on_worker_stop),
+                )
+                if inference_pool_gid
+                else self._default_pool
+            )
 
         env_hash = await compute_hash_of_file(env_tarball)
+        if inference_pool_gid is not None:
+            env_hash = _append_gid_environment_hash(env_hash, inference_pool_gid)
+
         if env_hash in self._pools:
@@ -133,8 +163,8 @@ return self._pools[env_hash]
         env = await self._extract_tarball(env_hash, env_tarball)
-        pool = InferencePool(
+        self._pools[env_hash] = InferencePool(
             self._settings, env=env, on_worker_stop=self._on_worker_stop
         )
-        self._pools[env_hash] = pool
-        return pool
 
+        return self._pools[env_hash]
+
     async def _extract_tarball(self, env_hash: str, env_tarball: str) -> Environment:
@@ -154,4 +184,13 @@ env_path = self._get_env_path(env_hash)
         env_hash = _get_environment_hash(model)
+        inference_pool_gid = (
+            model.settings.parameters.inference_pool_gid
+            if model.settings.parameters
+            else None
+        )
+
         if not env_hash:
-            return self._default_pool
+            if not inference_pool_gid:
+                return self._default_pool
+            else:
+                return self._pools[inference_pool_gid]
 
@@ -272,2 +311,4 @@ if env_hash not in self._pools:
         if env_hash:
+            # force calling __del__ on `Environment` to clean up
+            self._pools[env_hash]._env = None  # pylint: disable=protected-access
             del self._pools[env_hash]

mlserver/settings.py

 import sys
 import os
+import uuid
 import json
@@ -17,2 +18,3 @@ import importlib
 )
+from typing_extensions import Self
 from pydantic import (
@@ -23,2 +25,3 @@ ImportString,
 )
+from pydantic import model_validator
 from pydantic._internal._validators import import_string
@@ -318,2 +321,8 @@ import pydantic_settings
 
+    inference_pool_gid: Optional[str] = None
+    """Inference pool group id to be used to serve this model."""
+
+    autogenerate_inference_pool_gid: bool = False
+    """Flag to autogenerate the inference pool group id for this model."""
+
     format: Optional[str] = None
@@ -329,3 +338,9 @@ """Format of the model (only available on certain runtimes)."""
 
+    @model_validator(mode="after")
+    def set_inference_pool_gid(self) -> Self:
+        if self.autogenerate_inference_pool_gid and self.inference_pool_gid is None:
+            self.inference_pool_gid = str(uuid.uuid4())
+        return self
 
+
 class ModelSettings(BaseSettings):
@@ -332,0 +347,0 @@ model_config = SettingsConfigDict(

mlserver/version.py

@@ -1,1 +0,1 @@
-__version__ = "1.6.2.rc1"
+__version__ = "1.7.0.rc1"

PKG-INFO

@@ -1,4 +0,4 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.3
 Name: mlserver
-Version: 1.6.2rc1
+Version: 1.7.0rc1
 Summary: MLServer
@@ -8,3 +8,3 @@ License: Apache-2.0
 Author-email: [e-mail]
-Requires-Python: >=3.9,<3.12
+Requires-Python: >=3.9,<3.13
 Classifier: License :: OSI Approved :: Apache Software License
@@ -17,2 +17,3 @@ Classifier: Operating System :: MacOS
 Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
 Requires-Dist: aiofiles
@@ -36,3 +37,3 @@ Requires-Dist: aiokafka
 Requires-Dist: pydantic (>=2.7.1,<3.0.0)
-Requires-Dist: pydantic-settings (>=2.2.1,<3.0.0)
+Requires-Dist: pydantic-settings (>=2.3.0,<3.0.0)
 Requires-Dist: python-dotenv
@@ -148,4 +149,5 @@ Requires-Dist: python-multipart
 | 3.10           | 🟢     |
-| 3.11           | 🔵     |
-| 3.12           | 🔵     |
+| 3.11           | 🟢     |
+| 3.12           | 🟢     |
+| 3.13           | 🔴     |
 
@@ -152,0 +154,0 @@ ## Examples

pyproject.toml

 [tool.poetry]
 name = "mlserver"
-version = "1.6.2.rc1"
+version = "1.7.0.rc1"
 description = "MLServer"
@@ -46,3 +46,3 @@ authors = ["Seldon Technologies Ltd. <[e-mail]>"]
 [tool.poetry.dependencies]
-python = ">=3.9,<3.12"
+python = ">=3.9,<3.13"
 click = "*"
@@ -67,3 +67,3 @@ fastapi = ">=0.88.0,!=0.89.0,<0.116.0"
 pydantic = "^2.7.1"
-pydantic-settings = "^2.2.1"
+pydantic-settings = "^2.3.0"
 python-multipart = "*"
@@ -97,3 +97,3 @@
 httpx = "0.27.0"
-kafka-python = "2.0.2"
+kafka-python-ng = "2.2.3"
 tenacity = "8.4.1"
@@ -117,3 +117,3 @@ pyyaml = "6.0.1"
 [tool.poetry.group.docker.dependencies]
-tensorflow = "<2.15"
+tensorflow = "^2.16"
 
@@ -142,10 +142,12 @@ [tool.poetry.group.all-runtimes]
 ## Dev dependencies from MLflow
-torch = "2.2.1"
-pytorch-lightning = "2.4.0"
-torchmetrics = "1.5.0"
-torchvision = "0.17.1"
-mlflow = "2.17.0"
+torch = "^2.4"
+pytorch-lightning = "^2.4"
+torchmetrics = "1.6.0"
+torchvision = "0.19.1"
+mlflow = "2.19.0"
 
 ## Dev dependencies from HuggingFace
-transformers = ">=4.30,<5.0"
+# TODO: Relax when we deprecate Conversation pipeline
+# see: https://github.com/SeldonIO/MLServer/issues/1955
+transformers = "4.41.2"
 
@@ -152,0 +154,0 @@ [tool.poetry.group.docs]

README.md

@@ -103,4 +103,5 @@ # MLServer
 | 3.10           | 🟢     |
-| 3.11           | 🔵     |
-| 3.12           | 🔵     |
+| 3.11           | 🟢     |
+| 3.12           | 🟢     |
+| 3.13           | 🔴     |
 
@@ -107,0 +108,0 @@ ## Examples

Alert delta unavailable

Currently unable to show alert delta for PyPI packages.

Improved metrics

Total package byte prevSize

474890

1.15%

Lines of code

11510

1.14%