modal 0.75.7__py3-none-any.whl → 0.75.8__py3-none-any.whl

modal/_functions.py

@@ -1075,7 +1075,9 @@ class _Function(typing.Generic[P, ReturnType, OriginalReturnType], _Object, type
                 serialized_params = serialize_proto_params(kwargs_with_defaults)
                 can_use_parent = len(parent._class_parameter_info.schema) == 0  # no parameters
             else:
-                can_use_parent = len(args) + len(kwargs) == 0 and options is None
+                from modal.cls import _ServiceOptions  # Should probably define this dataclass here?
+
+                can_use_parent = len(args) + len(kwargs) == 0 and (options == _ServiceOptions())
                 serialized_params = serialize((args, kwargs))
 
             if can_use_parent:
@@ -1098,14 +1100,18 @@ class _Function(typing.Generic[P, ReturnType, OriginalReturnType], _Object, type
                 options_pb = api_pb2.FunctionOptions(
                     secret_ids=[s.object_id for s in options.secrets],
                     replace_secret_ids=bool(options.secrets),
+                    replace_volume_mounts=len(volume_mounts) > 0,
+                    volume_mounts=volume_mounts,
                     resources=options.resources,
                     retry_policy=options.retry_policy,
-                    concurrency_limit=options.concurrency_limit,
+                    concurrency_limit=options.max_containers,
+                    buffer_containers=options.buffer_containers,
+                    task_idle_timeout_secs=options.scaledown_window,
                     timeout_secs=options.timeout_secs,
-                    task_idle_timeout_secs=options.task_idle_timeout_secs,
-                    replace_volume_mounts=len(volume_mounts) > 0,
-                    volume_mounts=volume_mounts,
+                    max_concurrent_inputs=options.max_concurrent_inputs,
                     target_concurrent_inputs=options.target_concurrent_inputs,
+                    batch_max_size=options.batch_max_size,
+                    batch_linger_ms=options.batch_wait_ms,
                 )
             else:
                 options_pb = None
@@ -1399,7 +1405,7 @@ class _Function(typing.Generic[P, ReturnType, OriginalReturnType], _Object, type
             definition_id=self._definition_id,
             method_handle_metadata=self._method_handle_metadata,
             function_schema=self._metadata.function_schema if self._metadata else None,
-            input_plane_url=self._input_plane_url
+            input_plane_url=self._input_plane_url,
         )
 
     def _check_no_web_url(self, fn_name: str):

modal/client.pyi

@@ -27,7 +27,7 @@ class _Client:
     _snapshotted: bool
 
     def __init__(
-        self, server_url: str, client_type: int, credentials: typing.Optional[tuple[str, str]], version: str = "0.75.7"
+        self, server_url: str, client_type: int, credentials: typing.Optional[tuple[str, str]], version: str = "0.75.8"
     ): ...
     def is_closed(self) -> bool: ...
     @property
@@ -86,7 +86,7 @@ class Client:
     _snapshotted: bool
 
     def __init__(
-        self, server_url: str, client_type: int, credentials: typing.Optional[tuple[str, str]], version: str = "0.75.7"
+        self, server_url: str, client_type: int, credentials: typing.Optional[tuple[str, str]], version: str = "0.75.8"
     ): ...
     def is_closed(self) -> bool: ...
     @property

modal/cls.py

@@ -75,14 +75,18 @@ def _get_class_constructor_signature(user_cls: type) -> inspect.Signature:
 
 @dataclasses.dataclass()
 class _ServiceOptions:
-    secrets: typing.Collection[_Secret]
-    resources: Optional[api_pb2.Resources]
-    retry_policy: Optional[api_pb2.FunctionRetryPolicy]
-    concurrency_limit: Optional[int]
-    timeout_secs: Optional[int]
-    task_idle_timeout_secs: Optional[int]
-    validated_volumes: typing.Sequence[tuple[str, _Volume]]
-    target_concurrent_inputs: Optional[int]
+    secrets: typing.Collection[_Secret] = ()
+    validated_volumes: typing.Sequence[tuple[str, _Volume]] = ()
+    resources: Optional[api_pb2.Resources] = None
+    retry_policy: Optional[api_pb2.FunctionRetryPolicy] = None
+    max_containers: Optional[int] = None
+    buffer_containers: Optional[int] = None
+    scaledown_window: Optional[int] = None
+    timeout_secs: Optional[int] = None
+    max_concurrent_inputs: Optional[int] = None
+    target_concurrent_inputs: Optional[int] = None
+    batch_max_size: Optional[int] = None
+    batch_wait_ms: Optional[int] = None
 
 
 def _bind_instance_method(cls: "_Cls", service_function: _Function, method_name: str):
@@ -421,7 +425,7 @@ class _Cls(_Object, type_prefix="cs"):
     """
 
     _class_service_function: Optional[_Function]  # The _Function (read "service") serving *all* methods of the class
-    _options: Optional[_ServiceOptions]
+    _options: _ServiceOptions
 
     _app: Optional["modal.app._App"] = None  # not set for lookups
     _name: Optional[str]
@@ -437,7 +441,7 @@ class _Cls(_Object, type_prefix="cs"):
     def _initialize_from_empty(self):
         self._user_cls = None
         self._class_service_function = None
-        self._options = None
+        self._options = _ServiceOptions()
         self._callables = {}
         self._name = None
 
@@ -659,17 +663,15 @@ More information on class parameterization can be found here: https://modal.com/
         volumes: dict[Union[str, os.PathLike], _Volume] = {},
         retries: Optional[Union[int, Retries]] = None,
         max_containers: Optional[int] = None,  # Limit on the number of containers that can be concurrently running.
+        buffer_containers: Optional[int] = None,  # Additional containers to scale up while Function is active.
         scaledown_window: Optional[int] = None,  # Max amount of time a container can remain idle before scaling down.
         timeout: Optional[int] = None,
-        allow_concurrent_inputs: Optional[int] = None,
         # The following parameters are deprecated
         concurrency_limit: Optional[int] = None,  # Now called `max_containers`
         container_idle_timeout: Optional[int] = None,  # Now called `scaledown_window`
+        allow_concurrent_inputs: Optional[int] = None,  # See `.with_concurrency`
     ) -> "_Cls":
-        """
-        **Beta:** Allows for the runtime modification of a modal.Cls's configuration.
-
-        This is a beta feature and may be unstable.
+        """Create an instance of the Cls with configuration options overridden with new values.
 
         **Usage:**
 
@@ -685,6 +687,13 @@ More information on class parameterization can be found here: https://modal.com/
         else:
             resources = None
 
+        if allow_concurrent_inputs is not None:
+            deprecation_warning(
+                (2025, 5, 9),
+                "The `allow_concurrent_inputs` argument is deprecated;"
+                " please use the `.with_concurrency` method instead.",
+            )
+
         async def _load_from_base(new_cls, resolver, existing_object_id):
             # this is a bit confusing, the cls will always have the same metadata
             # since it has the same *class* service function (i.e. "template")
@@ -703,19 +712,79 @@ More information on class parameterization can be found here: https://modal.com/
 
         cls = _Cls._from_loader(_load_from_base, rep=f"{self._name}.with_options(...)", is_another_app=True, deps=_deps)
         cls._initialize_from_other(self)
-        cls._options = _ServiceOptions(
+        cls._options = dataclasses.replace(
+            cls._options,
             secrets=secrets,
             resources=resources,
             retry_policy=retry_policy,
-            # TODO(michael) Update the protos to use the new terminology
-            concurrency_limit=max_containers,
-            task_idle_timeout_secs=scaledown_window,
+            max_containers=max_containers,
+            buffer_containers=buffer_containers,
+            scaledown_window=scaledown_window,
             timeout_secs=timeout,
             validated_volumes=validate_volumes(volumes),
+            # Note: set both for backwards / forwards compatibility
+            # But going forward `.with_concurrency` is the preferred method with distinct parameterization
+            max_concurrent_inputs=allow_concurrent_inputs,
             target_concurrent_inputs=allow_concurrent_inputs,
         )
         return cls
 
+    def with_concurrency(self: "_Cls", *, max_inputs: int, target_inputs: Optional[int] = None) -> "_Cls":
+        """Create an instance of the Cls with input concurrency enabled or overridden with new values.
+
+        **Usage:**
+
+        ```python notest
+        Model = modal.Cls.from_name("my_app", "Model")
+        ModelUsingGPU = Model.with_options(gpu="A100").with_concurrency(max_inputs=100)
+        ModelUsingGPU().generate.remote(42)  # will run on an A100 GPU with input concurrency enabled
+        ```
+        """
+
+        async def _load_from_base(new_cls, resolver, existing_object_id):
+            if not self.is_hydrated:
+                await resolver.load(self)
+            new_cls._initialize_from_other(self)
+
+        def _deps():
+            return []
+
+        cls = _Cls._from_loader(
+            _load_from_base, rep=f"{self._name}.with_concurrency(...)", is_another_app=True, deps=_deps
+        )
+        cls._initialize_from_other(self)
+        cls._options = dataclasses.replace(
+            cls._options, max_concurrent_inputs=max_inputs, target_concurrent_inputs=target_inputs
+        )
+        return cls
+
+    def with_batching(self: "_Cls", *, max_batch_size: int, wait_ms: int) -> "_Cls":
+        """Create an instance of the Cls with dynamic batching enabled or overridden with new values.
+
+        **Usage:**
+
+        ```python notest
+        Model = modal.Cls.from_name("my_app", "Model")
+        ModelUsingGPU = Model.with_options(gpu="A100").with_batching(max_batch_size=100, batch_wait_ms=1000)
+        ModelUsingGPU().generate.remote(42)  # will run on an A100 GPU with input concurrency enabled
+        ```
+        """
+
+        async def _load_from_base(new_cls, resolver, existing_object_id):
+            if not self.is_hydrated:
+                await resolver.load(self)
+            new_cls._initialize_from_other(self)
+
+        def _deps():
+            return []
+
+        cls = _Cls._from_loader(
+            _load_from_base, rep=f"{self._name}.with_concurrency(...)", is_another_app=True, deps=_deps
+        )
+        cls._initialize_from_other(self)
+        cls._options = dataclasses.replace(cls._options, batch_max_size=max_batch_size, batch_wait_ms=wait_ms)
+        return cls
+
     @staticmethod
     @renamed_parameter((2024, 12, 18), "tag", "name")
     async def lookup(

modal/cls.pyi

@@ -25,24 +25,32 @@ def _get_class_constructor_signature(user_cls: type) -> inspect.Signature: ...
 
 class _ServiceOptions:
     secrets: typing.Collection[modal.secret._Secret]
+    validated_volumes: typing.Sequence[tuple[str, modal.volume._Volume]]
     resources: typing.Optional[modal_proto.api_pb2.Resources]
     retry_policy: typing.Optional[modal_proto.api_pb2.FunctionRetryPolicy]
-    concurrency_limit: typing.Optional[int]
+    max_containers: typing.Optional[int]
+    buffer_containers: typing.Optional[int]
+    scaledown_window: typing.Optional[int]
     timeout_secs: typing.Optional[int]
-    task_idle_timeout_secs: typing.Optional[int]
-    validated_volumes: typing.Sequence[tuple[str, modal.volume._Volume]]
+    max_concurrent_inputs: typing.Optional[int]
     target_concurrent_inputs: typing.Optional[int]
+    batch_max_size: typing.Optional[int]
+    batch_wait_ms: typing.Optional[int]
 
     def __init__(
         self,
-        secrets: typing.Collection[modal.secret._Secret],
-        resources: typing.Optional[modal_proto.api_pb2.Resources],
-        retry_policy: typing.Optional[modal_proto.api_pb2.FunctionRetryPolicy],
-        concurrency_limit: typing.Optional[int],
-        timeout_secs: typing.Optional[int],
-        task_idle_timeout_secs: typing.Optional[int],
-        validated_volumes: typing.Sequence[tuple[str, modal.volume._Volume]],
-        target_concurrent_inputs: typing.Optional[int],
+        secrets: typing.Collection[modal.secret._Secret] = (),
+        validated_volumes: typing.Sequence[tuple[str, modal.volume._Volume]] = (),
+        resources: typing.Optional[modal_proto.api_pb2.Resources] = None,
+        retry_policy: typing.Optional[modal_proto.api_pb2.FunctionRetryPolicy] = None,
+        max_containers: typing.Optional[int] = None,
+        buffer_containers: typing.Optional[int] = None,
+        scaledown_window: typing.Optional[int] = None,
+        timeout_secs: typing.Optional[int] = None,
+        max_concurrent_inputs: typing.Optional[int] = None,
+        target_concurrent_inputs: typing.Optional[int] = None,
+        batch_max_size: typing.Optional[int] = None,
+        batch_wait_ms: typing.Optional[int] = None,
     ) -> None: ...
     def __repr__(self): ...
     def __eq__(self, other): ...
@@ -141,7 +149,7 @@ class Obj:
 
 class _Cls(modal._object._Object):
     _class_service_function: typing.Optional[modal._functions._Function]
-    _options: typing.Optional[_ServiceOptions]
+    _options: _ServiceOptions
     _app: typing.Optional[modal.app._App]
     _name: typing.Optional[str]
     _method_metadata: typing.Optional[dict[str, modal_proto.api_pb2.FunctionHandleMetadata]]
@@ -182,12 +190,15 @@ class _Cls(modal._object._Object):
         volumes: dict[typing.Union[str, os.PathLike], modal.volume._Volume] = {},
         retries: typing.Union[int, modal.retries.Retries, None] = None,
         max_containers: typing.Optional[int] = None,
+        buffer_containers: typing.Optional[int] = None,
         scaledown_window: typing.Optional[int] = None,
         timeout: typing.Optional[int] = None,
-        allow_concurrent_inputs: typing.Optional[int] = None,
         concurrency_limit: typing.Optional[int] = None,
         container_idle_timeout: typing.Optional[int] = None,
+        allow_concurrent_inputs: typing.Optional[int] = None,
     ) -> _Cls: ...
+    def with_concurrency(self: _Cls, *, max_inputs: int, target_inputs: typing.Optional[int] = None) -> _Cls: ...
+    def with_batching(self: _Cls, *, max_batch_size: int, wait_ms: int) -> _Cls: ...
     @staticmethod
     async def lookup(
         app_name: str,
@@ -203,7 +214,7 @@ class _Cls(modal._object._Object):
 
 class Cls(modal.object.Object):
     _class_service_function: typing.Optional[modal.functions.Function]
-    _options: typing.Optional[_ServiceOptions]
+    _options: _ServiceOptions
     _app: typing.Optional[modal.app.App]
     _name: typing.Optional[str]
     _method_metadata: typing.Optional[dict[str, modal_proto.api_pb2.FunctionHandleMetadata]]
@@ -245,12 +256,15 @@ class Cls(modal.object.Object):
         volumes: dict[typing.Union[str, os.PathLike], modal.volume.Volume] = {},
         retries: typing.Union[int, modal.retries.Retries, None] = None,
         max_containers: typing.Optional[int] = None,
+        buffer_containers: typing.Optional[int] = None,
         scaledown_window: typing.Optional[int] = None,
         timeout: typing.Optional[int] = None,
-        allow_concurrent_inputs: typing.Optional[int] = None,
         concurrency_limit: typing.Optional[int] = None,
         container_idle_timeout: typing.Optional[int] = None,
+        allow_concurrent_inputs: typing.Optional[int] = None,
     ) -> Cls: ...
+    def with_concurrency(self: Cls, *, max_inputs: int, target_inputs: typing.Optional[int] = None) -> Cls: ...
+    def with_batching(self: Cls, *, max_batch_size: int, wait_ms: int) -> Cls: ...
 
     class __lookup_spec(typing_extensions.Protocol):
         def __call__(

{modal-0.75.7.dist-info → modal-0.75.8.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: modal
-Version: 0.75.7
+Version: 0.75.8
 Summary: Python client library for Modal
 Author-email: Modal Labs <support@modal.com>
 License: Apache-2.0

{modal-0.75.7.dist-info → modal-0.75.8.dist-info}/RECORD

@@ -3,7 +3,7 @@ modal/__main__.py,sha256=sTJcc9EbDuCKSwg3tL6ZckFw9WWdlkXW8mId1IvJCNc,2846
 modal/_clustered_functions.py,sha256=kTf-9YBXY88NutC1akI-gCbvf01RhMPCw-zoOI_YIUE,2700
 modal/_clustered_functions.pyi,sha256=2aWxN2v5WUnj-R-sk6BzJ-3AvggkQGQjwhtvbDH3pds,777
 modal/_container_entrypoint.py,sha256=2Zx9O_EMJg0H77EdnC2vGKs6uFMWwbP1NLFf-qYmWmU,28962
-modal/_functions.py,sha256=7Cec6Q7_ND6Ejlwn0c6Y2OufOmsg6huWqeRTZY6faYM,80527
+modal/_functions.py,sha256=MbH2KMEpzcurMnwUZijf8UStcEiUDEEUHvB2ZbhRYzQ,80892
 modal/_ipython.py,sha256=TW1fkVOmZL3YYqdS2YlM1hqpf654Yf8ZyybHdBnlhSw,301
 modal/_location.py,sha256=joiX-0ZeutEUDTrrqLF1GHXCdVLF-rHzstocbMcd_-k,366
 modal/_object.py,sha256=6ve4sI2nRAnjPCuAXdSoUplaXfzC9MqRlF_ZLULwwy0,11472
@@ -22,11 +22,11 @@ modal/app.py,sha256=xojuGZv4LaQwZU5ntj7WbmMjeNuB9Gll8Mzqe2LyiEs,51323
 modal/app.pyi,sha256=zNwR1_2LpmQc9AhemuAeVdk90XNYDw9keOkXAwAATeA,28732
 modal/call_graph.py,sha256=1g2DGcMIJvRy-xKicuf63IVE98gJSnQsr8R_NVMptNc,2581
 modal/client.py,sha256=o-aQThHpvDHUzg_kUafyhWzACViUBhY2WLZ2EitnSHA,16787
-modal/client.pyi,sha256=qw-zVt9oNSmGtZ4Crjb5Y3-RYABJZ5v0h9pXvvwpQvs,8383
+modal/client.pyi,sha256=xc5cIKiyJDEfzGWLoqKEFF6Vz3GV3nTHQfmSI-8G-uQ,8383
 modal/cloud_bucket_mount.py,sha256=YOe9nnvSr4ZbeCn587d7_VhE9IioZYRvF9VYQTQux08,5914
 modal/cloud_bucket_mount.pyi,sha256=30T3K1a89l6wzmEJ_J9iWv9SknoGqaZDx59Xs-ZQcmk,1607
-modal/cls.py,sha256=aHoMEWMZUN7bOezs3tRPxzS1FP3gTxZBORVjbPmtxyg,35338
-modal/cls.pyi,sha256=klQkXHXhl5gTjRo_XDV6PGecnJeFOudSMXid_kanumI,12090
+modal/cls.py,sha256=LZMQdFZ06LpbLDbsCJknniU-393Jb44E1lqveLVM8F8,38365
+modal/cls.pyi,sha256=BmFCoiFS4EQjkx36iybAFNpxu6xaq24lyJpQuCsB-ko,12990
 modal/config.py,sha256=OOMEJ5LHNFbHRW5wUpuhl0TH6EPW8D1XV9I3OJXrZrk,12668
 modal/container_process.py,sha256=vvyK3DVPUMsuqvkKdUiQ49cDLF9JawGrxpglLk5vfgI,6208
 modal/container_process.pyi,sha256=cR4aRHTbcVvmxGCc1_k2Ey8JllJIAQYq9PBKx0_1TuI,2916
@@ -146,7 +146,7 @@ modal/requirements/2024.10.txt,sha256=qD-5cVIVM9wXesJ6JC89Ew-3m2KjEElUz3jaw_MddR
 modal/requirements/PREVIEW.txt,sha256=qD-5cVIVM9wXesJ6JC89Ew-3m2KjEElUz3jaw_MddRo,296
 modal/requirements/README.md,sha256=9tK76KP0Uph7O0M5oUgsSwEZDj5y-dcUPsnpR0Sc-Ik,854
 modal/requirements/base-images.json,sha256=57vMSqzMbLBxw5tFWSaMiIkkVEps4JfX5PAtXGnkS4U,740
-modal-0.75.7.dist-info/licenses/LICENSE,sha256=psuoW8kuDP96RQsdhzwOqi6fyWv0ct8CR6Jr7He_P_k,10173
+modal-0.75.8.dist-info/licenses/LICENSE,sha256=psuoW8kuDP96RQsdhzwOqi6fyWv0ct8CR6Jr7He_P_k,10173
 modal_docs/__init__.py,sha256=svYKtV8HDwDCN86zbdWqyq5T8sMdGDj0PVlzc2tIxDM,28
 modal_docs/gen_cli_docs.py,sha256=c1yfBS_x--gL5bs0N4ihMwqwX8l3IBWSkBAKNNIi6bQ,3801
 modal_docs/gen_reference_docs.py,sha256=d_CQUGQ0rfw28u75I2mov9AlS773z9rG40-yq5o7g2U,6359
@@ -171,9 +171,9 @@ modal_proto/options_pb2_grpc.pyi,sha256=CImmhxHsYnF09iENPoe8S4J-n93jtgUYD2JPAc0y
 modal_proto/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 modal_version/__init__.py,sha256=PenIvZdwt-HVdbetAyxuPoyZTtzx2moKZoJLK8iZ804,470
 modal_version/__main__.py,sha256=2FO0yYQQwDTh6udt1h-cBnGd1c4ZyHnHSI4BksxzVac,105
-modal_version/_version_generated.py,sha256=6BcCuYUedZHHqWpjNiOKZgZruTK1fa-CqRO57DLf9yw,148
-modal-0.75.7.dist-info/METADATA,sha256=YyRzyMBDSH3pwd0JXc_xUh6CfzWSlOA7UOdYa7prIAQ,2450
-modal-0.75.7.dist-info/WHEEL,sha256=1tXe9gY0PYatrMPMDd6jXqjfpz_B-Wqm32CPfRC58XU,91
-modal-0.75.7.dist-info/entry_points.txt,sha256=An-wYgeEUnm6xzrAP9_NTSTSciYvvEWsMZILtYrvpAI,46
-modal-0.75.7.dist-info/top_level.txt,sha256=4BWzoKYREKUZ5iyPzZpjqx4G8uB5TWxXPDwibLcVa7k,43
-modal-0.75.7.dist-info/RECORD,,
+modal_version/_version_generated.py,sha256=snRSmP2yGv1hMbkNaSvYt_zoCWRhCWupozd1USbCSRE,148
+modal-0.75.8.dist-info/METADATA,sha256=PSEEBisDZSgqxMFVHeckk0BIUYsGRuRLMLXpLBS01i4,2450
+modal-0.75.8.dist-info/WHEEL,sha256=1tXe9gY0PYatrMPMDd6jXqjfpz_B-Wqm32CPfRC58XU,91
+modal-0.75.8.dist-info/entry_points.txt,sha256=An-wYgeEUnm6xzrAP9_NTSTSciYvvEWsMZILtYrvpAI,46
+modal-0.75.8.dist-info/top_level.txt,sha256=4BWzoKYREKUZ5iyPzZpjqx4G8uB5TWxXPDwibLcVa7k,43
+modal-0.75.8.dist-info/RECORD,,

modal_version/_version_generated.py

@@ -1,4 +1,4 @@
 # Copyright Modal Labs 2025
 
 # Note: Reset this value to -1 whenever you make a minor `0.X` release of the client.
-build_number = 7  # git: ebed27f
+build_number = 8  # git: 75ded48