python/tvm_ffi/_convert.py - tvm-ffi - Git at Google

 # Licensed to the Apache Software Foundation (ASF) under one
 # or more contributor license agreements.  See the NOTICE file
 # distributed with this work for additional information
 # regarding copyright ownership.  The ASF licenses this file
 # to you under the Apache License, Version 2.0 (the
 # "License"); you may not use this file except in compliance
 # with the License.  You may obtain a copy of the License at
 #
 #   http://www.apache.org/licenses/LICENSE-2.0
 #
 # Unless required by applicable law or agreed to in writing,
 # software distributed under the License is distributed on an
 # "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
 # KIND, either express or implied.  See the License for the
 # specific language governing permissions and limitations
 # under the License.
 """Conversion utilities to convert Python objects into TVM FFI values."""

 from __future__ import annotations

 import ctypes
 from numbers import Number
 from types import ModuleType
 from typing import Any, Callable

 from . import _dtype, container, core

 try:
     import torch
 except ImportError:
     torch = None  # ty: ignore[invalid-assignment]

 numpy: ModuleType | None = None
 try:
     import numpy
 except ImportError:
     pass


 def convert(value: Any) -> Any:  # noqa: PLR0911,PLR0912
     """Convert a Python object into TVM FFI values.

     This helper mirrors the automatic argument conversion that happens when
     calling FFI functions. It is primarily useful in tests or places where
     an explicit conversion is desired.

     Parameters
     ----------
     value
         The Python object to be converted.

     Returns
     -------
     ffi_obj
         The converted TVM FFI object.

     Examples
     --------
     .. code-block:: python

         import tvm_ffi

         # Lists and tuples become tvm_ffi.Array
         a = tvm_ffi.convert([1, 2, 3])
         assert isinstance(a, tvm_ffi.Array)

         # Dicts become tvm_ffi.Map
         m = tvm_ffi.convert({"a": 1, "b": 2})
         assert isinstance(m, tvm_ffi.Map)

         # Strings and bytes become zero-copy FFI-aware types
         s = tvm_ffi.convert("hello")
         b = tvm_ffi.convert(b"bytes")
         assert isinstance(s, tvm_ffi.core.String)
         assert isinstance(b, tvm_ffi.core.Bytes)

         # Callables are wrapped as tvm_ffi.Function
         f = tvm_ffi.convert(lambda x: x + 1)
         assert isinstance(f, tvm_ffi.Function)

         # Array libraries that support DLPack export can be converted to Tensor
         import numpy as np

         x = tvm_ffi.convert(np.arange(4, dtype="int32"))
         assert isinstance(x, tvm_ffi.Tensor)

     Note
     ----
     Function arguments to ffi function calls are
     automatically converted. So this function is mainly
     only used in internal or testing scenarios.

     """
     if isinstance(
         value, (core.Object, core.PyNativeObject, bool, Number, ctypes.c_void_p, _dtype.dtype)
     ):
         return value
     elif isinstance(value, (tuple, list)):
         return container.Array(value)
     elif isinstance(value, dict):
         return container.Map(value)
     elif isinstance(value, str):
         return core.String(value)
     elif isinstance(value, (bytes, bytearray)):
         return core.Bytes(value)
     elif isinstance(value, core.ObjectConvertible):
         return value.asobject()
     elif callable(value):
         return core._convert_to_ffi_func(value)
     elif value is None:
         return None
     elif hasattr(value, "__dlpack__"):
         return core.from_dlpack(value)
     elif torch is not None and isinstance(value, torch.dtype):
         return core._convert_torch_dtype_to_ffi_dtype(value)
     elif numpy is not None and isinstance(value, numpy.dtype):
         return core._convert_numpy_dtype_to_ffi_dtype(value)
     elif hasattr(value, "__dlpack_data_type__"):
         cdtype = core._create_cdtype_from_tuple(core.DataType, *value.__dlpack_data_type__())
         dtype = str.__new__(_dtype.dtype, str(cdtype))
         dtype._tvm_ffi_dtype = cdtype
         return dtype
     elif isinstance(value, Exception):
         return core._convert_to_ffi_error(value)
     elif hasattr(value, "__tvm_ffi_object__"):
         return value.__tvm_ffi_object__()
     # keep rest protocol values as it is as they can be handled by ffi function
     elif hasattr(value, "__cuda_stream__"):
         return value
     elif hasattr(value, "__tvm_ffi_opaque_ptr__"):
         return value
     elif hasattr(value, "__dlpack_device__"):
         return value
     elif hasattr(value, "__tvm_ffi_int__"):
         return value
     elif hasattr(value, "__tvm_ffi_float__"):
         return value
     else:
         # in this case, it is an opaque python object
         return core._convert_to_opaque_object(value)


 def convert_func(
     pyfunc: Callable[..., Any],
     tensor_cls: type | None = None,
 ) -> Any:
     """Convert a Python callable to an FFI :py:class:`~tvm_ffi.Function`.

     This is the callable-specific sibling of :py:func:`tvm_ffi.convert`.
     It accepts one extra argument, ``tensor_cls``, that lets the caller
     specify how tensor arguments should be delivered to the Python
     callable when the resulting :py:class:`Function` is invoked from C++.
     :py:func:`tvm_ffi.convert` has no such knob — it always produces a
     :py:class:`Function` whose callback receives ``tvm_ffi.Tensor`` for
     tensor args.

     Parameters
     ----------
     pyfunc : Callable
         The Python callable to wrap.
     tensor_cls : type, optional
         The class whose instances the callback should receive for tensor
         args. The class must expose a ``__dlpack_c_exchange_api__``
         :py:class:`PyCapsule`; its capsule is threaded into the callback
         closure so tensor args are converted at the C level (via the
         DLPack exchange API) before the Python callback body runs — this
         is significantly faster than calling ``torch.from_dlpack(x)`` (or
         equivalent) inside the callback. Raises :py:class:`TypeError` if
         ``tensor_cls`` does not expose the attribute.

         When ``tensor_cls`` is ``None``, ``convert_func`` behaves like the
         callable branch of :py:func:`tvm_ffi.convert`.

     Returns
     -------
     Function
         The wrapped FFI function.

     Examples
     --------
     .. code-block:: python

         import torch
         import tvm_ffi

         # Without tensor_cls: same as tvm_ffi.convert(pyfunc) — the callback
         # receives tvm_ffi.Tensor for tensor args.
         f = tvm_ffi.convert_func(lambda x: x + 1)
         assert isinstance(f, tvm_ffi.Function)


         # With tensor_cls=torch.Tensor: the callback receives torch.Tensor
         # directly; the DLPack conversion happens in C before the body runs.
         def callback(a: torch.Tensor, b: torch.Tensor) -> torch.Tensor:
             return a + b


         g = tvm_ffi.convert_func(callback, tensor_cls=torch.Tensor)

     See Also
     --------
     :py:func:`tvm_ffi.convert` :
         Generic value-to-FFI conversion. Use this when you don't need to
         specify ``tensor_cls``.

     """
     return core._convert_to_ffi_func(pyfunc, tensor_cls=tensor_cls)
	# Licensed to the Apache Software Foundation (ASF) under one
	# or more contributor license agreements. See the NOTICE file
	# distributed with this work for additional information
	# regarding copyright ownership. The ASF licenses this file
	# to you under the Apache License, Version 2.0 (the
	# "License"); you may not use this file except in compliance
	# with the License. You may obtain a copy of the License at
	#
	# http://www.apache.org/licenses/LICENSE-2.0
	#
	# Unless required by applicable law or agreed to in writing,
	# software distributed under the License is distributed on an
	# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
	# KIND, either express or implied. See the License for the
	# specific language governing permissions and limitations
	# under the License.
	"""Conversion utilities to convert Python objects into TVM FFI values."""

	from __future__ import annotations

	import ctypes
	from numbers import Number
	from types import ModuleType
	from typing import Any, Callable

	from . import _dtype, container, core

	try:
	import torch
	except ImportError:
	torch = None # ty: ignore[invalid-assignment]

	numpy: ModuleType \| None = None
	try:
	import numpy
	except ImportError:
	pass


	def convert(value: Any) -> Any: # noqa: PLR0911,PLR0912
	"""Convert a Python object into TVM FFI values.

	This helper mirrors the automatic argument conversion that happens when
	calling FFI functions. It is primarily useful in tests or places where
	an explicit conversion is desired.

	Parameters
	----------
	value
	The Python object to be converted.

	Returns
	-------
	ffi_obj
	The converted TVM FFI object.

	Examples
	--------
	.. code-block:: python

	import tvm_ffi

	# Lists and tuples become tvm_ffi.Array
	a = tvm_ffi.convert([1, 2, 3])
	assert isinstance(a, tvm_ffi.Array)

	# Dicts become tvm_ffi.Map
	m = tvm_ffi.convert({"a": 1, "b": 2})
	assert isinstance(m, tvm_ffi.Map)

	# Strings and bytes become zero-copy FFI-aware types
	s = tvm_ffi.convert("hello")
	b = tvm_ffi.convert(b"bytes")
	assert isinstance(s, tvm_ffi.core.String)
	assert isinstance(b, tvm_ffi.core.Bytes)

	# Callables are wrapped as tvm_ffi.Function
	f = tvm_ffi.convert(lambda x: x + 1)
	assert isinstance(f, tvm_ffi.Function)

	# Array libraries that support DLPack export can be converted to Tensor
	import numpy as np

	x = tvm_ffi.convert(np.arange(4, dtype="int32"))
	assert isinstance(x, tvm_ffi.Tensor)

	Note
	----
	Function arguments to ffi function calls are
	automatically converted. So this function is mainly
	only used in internal or testing scenarios.

	"""
	if isinstance(
	value, (core.Object, core.PyNativeObject, bool, Number, ctypes.c_void_p, _dtype.dtype)
	):
	return value
	elif isinstance(value, (tuple, list)):
	return container.Array(value)
	elif isinstance(value, dict):
	return container.Map(value)
	elif isinstance(value, str):
	return core.String(value)
	elif isinstance(value, (bytes, bytearray)):
	return core.Bytes(value)
	elif isinstance(value, core.ObjectConvertible):
	return value.asobject()
	elif callable(value):
	return core._convert_to_ffi_func(value)
	elif value is None:
	return None
	elif hasattr(value, "__dlpack__"):
	return core.from_dlpack(value)
	elif torch is not None and isinstance(value, torch.dtype):
	return core._convert_torch_dtype_to_ffi_dtype(value)
	elif numpy is not None and isinstance(value, numpy.dtype):
	return core._convert_numpy_dtype_to_ffi_dtype(value)
	elif hasattr(value, "__dlpack_data_type__"):
	cdtype = core._create_cdtype_from_tuple(core.DataType, *value.__dlpack_data_type__())
	dtype = str.__new__(_dtype.dtype, str(cdtype))
	dtype._tvm_ffi_dtype = cdtype
	return dtype
	elif isinstance(value, Exception):
	return core._convert_to_ffi_error(value)
	elif hasattr(value, "__tvm_ffi_object__"):
	return value.__tvm_ffi_object__()
	# keep rest protocol values as it is as they can be handled by ffi function
	elif hasattr(value, "__cuda_stream__"):
	return value
	elif hasattr(value, "__tvm_ffi_opaque_ptr__"):
	return value
	elif hasattr(value, "__dlpack_device__"):
	return value
	elif hasattr(value, "__tvm_ffi_int__"):
	return value
	elif hasattr(value, "__tvm_ffi_float__"):
	return value
	else:
	# in this case, it is an opaque python object
	return core._convert_to_opaque_object(value)


	def convert_func(
	pyfunc: Callable[..., Any],
	tensor_cls: type \| None = None,
	) -> Any:
	"""Convert a Python callable to an FFI :py:class:`~tvm_ffi.Function`.

	This is the callable-specific sibling of :py:func:`tvm_ffi.convert`.
	It accepts one extra argument, ``tensor_cls``, that lets the caller
	specify how tensor arguments should be delivered to the Python
	callable when the resulting :py:class:`Function` is invoked from C++.
	:py:func:`tvm_ffi.convert` has no such knob — it always produces a
	:py:class:`Function` whose callback receives ``tvm_ffi.Tensor`` for
	tensor args.

	Parameters
	----------
	pyfunc : Callable
	The Python callable to wrap.
	tensor_cls : type, optional
	The class whose instances the callback should receive for tensor
	args. The class must expose a ``__dlpack_c_exchange_api__``
	:py:class:`PyCapsule`; its capsule is threaded into the callback
	closure so tensor args are converted at the C level (via the
	DLPack exchange API) before the Python callback body runs — this
	is significantly faster than calling ``torch.from_dlpack(x)`` (or
	equivalent) inside the callback. Raises :py:class:`TypeError` if
	``tensor_cls`` does not expose the attribute.

	When ``tensor_cls`` is ``None``, ``convert_func`` behaves like the
	callable branch of :py:func:`tvm_ffi.convert`.

	Returns
	-------
	Function
	The wrapped FFI function.

	Examples
	--------
	.. code-block:: python

	import torch
	import tvm_ffi

	# Without tensor_cls: same as tvm_ffi.convert(pyfunc) — the callback
	# receives tvm_ffi.Tensor for tensor args.
	f = tvm_ffi.convert_func(lambda x: x + 1)
	assert isinstance(f, tvm_ffi.Function)


	# With tensor_cls=torch.Tensor: the callback receives torch.Tensor
	# directly; the DLPack conversion happens in C before the body runs.
	def callback(a: torch.Tensor, b: torch.Tensor) -> torch.Tensor:
	return a + b


	g = tvm_ffi.convert_func(callback, tensor_cls=torch.Tensor)

	See Also
	--------
	:py:func:`tvm_ffi.convert` :
	Generic value-to-FFI conversion. Use this when you don't need to
	specify ``tensor_cls``.

	"""
	return core._convert_to_ffi_func(pyfunc, tensor_cls=tensor_cls)