spaces 0.50.dev2__tar.gz → 0.50.1__tar.gz

{spaces-0.50.dev2 → spaces-0.50.1}/PKG-INFO

@@ -1,20 +1,19 @@
 Metadata-Version: 2.4
 Name: spaces
-Version: 0.50.dev2
+Version: 0.50.1
 Summary: Utilities for Hugging Face Spaces
 Author: Charles Bensimon
 Author-email: Charles Bensimon <charles@huggingface.co>
 License-Expression: Apache-2.0
 Requires-Dist: packaging
 Requires-Dist: typing-extensions>=4,<5
+Requires-Dist: requests>=2.19,<3
+Requires-Dist: gradio
 Requires-Dist: pydantic>=1,<3
 Requires-Dist: httpx>=0.20
-Requires-Dist: fastapi ; extra == 'serving'
-Requires-Dist: uvicorn ; extra == 'serving'
 Requires-Python: >=3.10, <3.15
 Project-URL: Homepage, https://huggingface.co
 Project-URL: Repository, https://github.com/huggingface/huggingface_hub
-Provides-Extra: serving
 Description-Content-Type: text/markdown
 
 # Hugging Face Spaces

{spaces-0.50.dev2 → spaces-0.50.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "spaces"
-version = "0.50.dev2"
+version = "0.50.1"
 description = "Utilities for Hugging Face Spaces"
 readme = "README.md"
 requires-python = ">=3.10,<3.15"
@@ -11,16 +11,12 @@ authors = [
 dependencies = [
     "packaging",
     "typing-extensions>=4,<5",
+    "requests>=2.19,<3",
+    "gradio", # Actually requires 3.2+ if for `@spaces.GPU`
     "pydantic>=1,<3",
     "httpx>=0.20",
 ]
 
-[project.optional-dependencies]
-serving = [
-    "fastapi",
-    "uvicorn"
-]
-
 [project.urls]
 Homepage = "https://huggingface.co"
 Repository = "https://github.com/huggingface/huggingface_hub"
@@ -30,20 +26,20 @@ dev = [
     "pytest>=7.3.1,<8",
     "pytest-dotenv>=0.5.2,<0.6",
     "pytest-mock>=3.10.0,<4",
+    "requests-mock>=1.10.0,<2",
     "pytest-timeout>=2.1.0,<3",
     "poethepoet>=0.20.0,<0.21",
     "coverage>=7.2.6,<8",
     "pyright>=1.1.403,<2",
     "pytest-order>=1.1.0,<2",
     "pynvml>=11.5.0,<12",
+    "websockets>=13.0,<14",
     "pytest-asyncio>=0.21.1,<0.22",
     "pytest-httpx>=0.27.0,<0.28",
     "pydantic-to-typescript>=2.0.0,<3",
-    "requests>=2.19,<3",
     "setuptools",
 ]
 test = [
-    "gradio",
     "diffusers",
     "transformers",
     "accelerate",
@@ -55,13 +51,6 @@ test = [
     "kernels>=0.10.2,<0.11",
 ]
 
-[tool.uv]
-constraint-dependencies = [
-    "numpy<2",
-    "peft<0.19",
-    "transformers<4.51",
-]
-
 [tool.pytest.ini_options]
 addopts = "-v"
 testpaths = ["tests"]
@@ -71,6 +60,7 @@ env_files = ["tests/.env"]
 source = ["spaces"]
 omit = [
     "spaces/zero/api.py",
+    "spaces/zero/torch/aoti.py", # For now
     "spaces/zero/torch/patching_legacy.py",
     "spaces/zero/torch/bitsandbytes_legacy.py",
     "spaces/_vendor/*",
@@ -93,8 +83,6 @@ shell = """set -e
     coverage run -m pytest tests/cpu
     coverage run -m pytest tests/gpu
     find e2e -name '*.py' | xargs -I {} coverage run {}
-    pytest tests/cpu/serving/test_middleware.py::test_fds_server_hang --fd-xfail
-    pytest tests/cpu/serving/test_middleware.py::test_fds_requests_hang --fd-xfail
     coverage combine -q
     coverage xml  || true
     coverage html || true
@@ -113,11 +101,17 @@ shell = """set -e
 include = ["spaces", "tests"]
 exclude = ["spaces/_vendor"]
 pythonVersion = "3.10"
-reportMatchNotExhaustive = "error"
 reportPossiblyUnboundVariable = "error"
 reportUnnecessaryComparison = true
 reportPrivateImportUsage = "information"
 
+[tool.uv]
+constraint-dependencies = [
+    "numpy<2",
+    "peft<0.19",
+    "transformers<4.51",
+]
+
 [build-system]
 requires = ["uv_build>=0.11.6,<0.12"]
 build-backend = "uv_build"

{spaces-0.50.dev2 → spaces-0.50.1}/spaces/__init__.py

@@ -1,14 +1,12 @@
 """
 """
-
 import sys
-import warnings
 from types import FunctionType
 from typing import TYPE_CHECKING
 from typing import Callable
 
 
-# Prevent partially-imported gradio from importing spaces
+# Prevent gradio from importing spaces
 if (gr := sys.modules.get('gradio')) is not None: # pragma: no cover
     try:
         gr.Blocks
@@ -17,16 +15,9 @@ if (gr := sys.modules.get('gradio')) is not None: # pragma: no cover
 
 
 from .zero.decorator import GPU
-
-
-def gradio_auto_wrap(*args, **kwargs): # pragma: no cover
-    warnings.warn("gradio_auto_wrap has no effect and is deprecated")
-
-def disable_gradio_auto_wrap(*args, **kwargs): # pragma: no cover
-    warnings.warn("gradio_auto_wrap has no effect and is deprecated")
-
-def enable_gradio_auto_wrap(*args, **kwargs): # pragma: no cover
-    warnings.warn("gradio_auto_wrap has no effect and is deprecated")
+from .gradio import gradio_auto_wrap
+from .gradio import disable_gradio_auto_wrap
+from .gradio import enable_gradio_auto_wrap
 
 
 class LazyImported:
@@ -59,10 +50,30 @@ def _aoti_compile():
     from .zero.torch.aoti import aoti_compile
     return aoti_compile
 
+def _aoti_compile_and_save(): # pragma: no cover
+    from .zero.torch.aoti import aoti_compile_and_save
+    return aoti_compile_and_save
+
 def _aoti_apply():
     from .zero.torch.aoti import aoti_apply
     return aoti_apply
 
+def _aoti_patch(): # pragma: no cover
+    from .zero.torch.aoti import aoti_patch
+    return aoti_patch
+
+def _aoti_load(): # pragma: no cover
+    from .zero.torch.aoti import aoti_load
+    return aoti_load
+
+def _aoti_load_call_source(): # pragma: no cover
+    from .zero.torch.aoti import aoti_load_call_source
+    return aoti_load_call_source
+
+def _aoti_load_from_package_dir(): # pragma: no cover
+    from .zero.torch.aoti import aoti_load_from_package_dir
+    return aoti_load_from_package_dir
+
 def _aoti_blocks_load():
     from .zero.torch.aoti import aoti_blocks_load
     return aoti_blocks_load
@@ -70,14 +81,24 @@ def _aoti_blocks_load():
 
 aoti_capture = LazyImported(_aoti_capture)
 aoti_compile = LazyImported(_aoti_compile)
+aoti_compile_and_save = LazyImported(_aoti_compile_and_save)
 aoti_apply = LazyImported(_aoti_apply)
+aoti_patch = LazyImported(_aoti_patch)
+aoti_load = LazyImported(_aoti_load)
+aoti_load_call_source = LazyImported(_aoti_load_call_source)
+aoti_load_from_package_dir = LazyImported(_aoti_load_from_package_dir)
 aoti_blocks_load = LazyImported(_aoti_blocks_load)
 
 
 if TYPE_CHECKING:
     from .zero.torch.aoti import aoti_capture
     from .zero.torch.aoti import aoti_compile
+    from .zero.torch.aoti import aoti_compile_and_save
     from .zero.torch.aoti import aoti_apply
+    from .zero.torch.aoti import aoti_patch
+    from .zero.torch.aoti import aoti_load
+    from .zero.torch.aoti import aoti_load_call_source
+    from .zero.torch.aoti import aoti_load_from_package_dir
     from .zero.torch.aoti import aoti_blocks_load
 
 
@@ -88,6 +109,11 @@ __all__ = [
     'enable_gradio_auto_wrap',
     'aoti_capture',
     'aoti_compile',
+    'aoti_compile_and_save',
     'aoti_apply',
+    'aoti_patch',
+    'aoti_load',
+    'aoti_load_call_source',
+    'aoti_load_from_package_dir',
     'aoti_blocks_load',
 ]

{spaces-0.50.dev2 → spaces-0.50.1}/spaces/config.py

@@ -18,6 +18,8 @@ class Settings:
             os.getenv('SPACES_ZERO_GPU'))
         self.zero_device_api_url = (
             os.getenv('SPACES_ZERO_DEVICE_API_URL'))
+        self.gradio_auto_wrap = boolean(
+            os.getenv('SPACES_GRADIO_AUTO_WRAP'))
         self.zero_patch_torch_device = boolean(
             os.getenv('ZERO_GPU_PATCH_TORCH_DEVICE'))
         self.zero_gpu_v2 = boolean(

spaces-0.50.1/spaces/gradio.py

@@ -0,0 +1,53 @@
+"""
+"""
+from typing import Callable
+from typing import Generator
+from typing import TypeVar
+from typing import overload
+from typing_extensions import ParamSpec
+
+from .config import Config
+from .zero.decorator import GPU
+
+
+Param = ParamSpec('Param')
+Res = TypeVar('Res')
+
+
+gradio_auto_wrap_enabled = Config.gradio_auto_wrap
+
+
+def disable_gradio_auto_wrap():
+    global gradio_auto_wrap_enabled
+    gradio_auto_wrap_enabled = False
+
+def enable_gradio_auto_wrap():
+    global gradio_auto_wrap_enabled
+    gradio_auto_wrap_enabled = True
+
+
+@overload
+def gradio_auto_wrap(
+    task:
+     Callable[Param, Res],
+) -> Callable[Param, Res]:
+    ...
+@overload
+def gradio_auto_wrap(
+    task:
+     None,
+) -> None:
+    ...
+def gradio_auto_wrap(
+    task:
+      Callable[Param, Res]
+    | None,
+) -> (Callable[Param, Res]
+    | None):
+    """
+    """
+    if not gradio_auto_wrap_enabled:
+        return task
+    if not callable(task):
+        return task
+    return GPU(task) # type: ignore

{spaces-0.50.dev2 → spaces-0.50.1}/spaces/reloading/__init__.py

@@ -1,10 +1,8 @@
 """
 """
-
 import threading
 from typing import Any
 from typing import Callable
-from uuid import uuid4
 
 
 reload_server_ports: set[int] = set()
@@ -16,7 +14,6 @@ def start_reload_server(
     stop_event: threading.Event,
     port: int = 7887,
 ):
-    from ..zero.context import GlobalContext
     from .server import ReloadServer
 
     reload_server = ReloadServer(
@@ -25,9 +22,8 @@ def start_reload_server(
         stop_event=stop_event,
     )
 
-    server_port_uuid = uuid4()
-    GlobalContext.server_ports[server_port_uuid] = port
+    reload_server_ports.add(port)
     try:
         reload_server.run(port=port)
     finally:
-        del GlobalContext.server_ports[server_port_uuid]
+        reload_server_ports.remove(port)

{spaces-0.50.dev2 → spaces-0.50.1}/spaces/reloading/server.py

@@ -12,11 +12,6 @@ from typing_extensions import assert_type
 from typing_extensions import Never
 from uuid import uuid4
 
-try:
-    import fastapi, uvicorn
-except ModuleNotFoundError: # pragma: no cover
-    raise RuntimeError("spaces.reloading.server requires FastAPI and uvicorn")
-
 from fastapi import APIRouter
 from fastapi import FastAPI
 from fastapi import HTTPException

{spaces-0.50.dev2 → spaces-0.50.1}/spaces/utils.py

@@ -30,8 +30,10 @@ class SimpleQueue(_SimpleQueue[T]):
             raise # pragma: no cover
         # https://bugs.python.org/issue29187
         except Exception as e:
-            # Always assume a PicklingError
-            raise PicklingError from e
+            message = str(e)
+            if not "pickle" in message:
+                raise # pragma: no cover
+            raise PicklingError(message)
     def wlock_release(self):
         if (lock := getattr(self, '_wlock', None)) is None:
             return # pragma: no cover
@@ -47,12 +49,17 @@ def drop_params(fn: Callable[[], T]) -> Callable[..., T]:
     return drop
 
 
+def gradio_request_var():
+    try:
+        from gradio.context import LocalContext
+    except ImportError: # pragma: no cover
+        raise RuntimeError(GRADIO_VERSION_ERROR_MESSAGE)
+    return LocalContext.request
+
+
 debug = partial(print, 'SPACES_ZERO_GPU_DEBUG')
 
 
 # Type-safe threads
 def create_thread(fn: Callable[P, Any], *args: P.args, **kwargs: P.kwargs) -> Thread:
     return Thread(target=fn, args=args, kwargs=kwargs)
-
-def create_thread_daemon(fn: Callable[P, Any], *args: P.args, **kwargs: P.kwargs) -> Thread:
-    return Thread(target=fn, args=args, kwargs=kwargs, daemon=True)

spaces-0.50.1/spaces/zero/__init__.py

@@ -0,0 +1,41 @@
+"""
+"""
+
+from ..config import Config
+
+
+_default_gpu_size: 'api.GPUSize | None' = None
+
+
+def default_gpu_size():
+    return _default_gpu_size
+
+
+if Config.zero_gpu:
+
+    from . import api
+    from . import client
+    from . import config
+    from . import decorator
+    from . import gradio
+    from . import torch
+
+    if torch.is_in_bad_fork():
+        raise RuntimeError(
+            "CUDA has been initialized before importing the `spaces` package. "
+            "Try importing `spaces` before any other CUDA-related package."
+        )
+
+    def startup():
+        global _default_gpu_size
+        zerogpu_config = config.get_config()
+        total_size = torch.pack()
+        threshold = zerogpu_config['xlarge_threshold']
+        if threshold is not None and total_size > threshold:
+            _default_gpu_size = 'xlarge' # pragma: no cover
+        if len(decorator.decorated_cache) == 0:
+            return # pragma: no cover
+        client.startup_report()
+
+    torch.patch()
+    gradio.one_launch(startup)