gstaichi 0.1.18.dev1__cp310-cp310-win_amd64.whl

taichi/lang/field.py

@@ -0,0 +1,464 @@
+# type: ignore
+
+import taichi.lang
+from taichi._lib import core as _ti_core
+from taichi._logging import warn
+from taichi.lang import impl
+from taichi.lang.exception import TaichiSyntaxError
+from taichi.lang.util import (
+    in_python_scope,
+    python_scope,
+    to_numpy_type,
+    to_paddle_type,
+    to_pytorch_type,
+)
+
+
+class Field:
+    """Taichi field class.
+
+    A field is constructed by a list of field members.
+    For example, a scalar field has 1 field member, while a 3x3 matrix field has 9 field members.
+    A field member is a Python Expr wrapping a C++ FieldExpression.
+
+    Args:
+        vars (List[Expr]): Field members.
+    """
+
+    def __init__(self, _vars):
+        assert all(_vars)
+        self.vars = _vars
+        self.host_accessors = None
+        self.grad = None
+        self.dual = None
+
+    @property
+    def snode(self):
+        """Gets representative SNode for info purposes.
+
+        Returns:
+            SNode: Representative SNode (SNode of first field member).
+        """
+        return self._snode
+
+    @property
+    def _snode(self):
+        """Gets representative SNode for info purposes.
+
+        Returns:
+            SNode: Representative SNode (SNode of first field member).
+        """
+        return taichi.lang.snode.SNode(self.vars[0].ptr.snode())
+
+    @property
+    def shape(self):
+        """Gets field shape.
+
+        Returns:
+            Tuple[Int]: Field shape.
+        """
+        return self._snode.shape
+
+    @property
+    def dtype(self):
+        """Gets data type of each individual value.
+
+        Returns:
+            DataType: Data type of each individual value.
+        """
+        return self._snode._dtype
+
+    @property
+    def _name(self):
+        """Gets field name.
+
+        Returns:
+            str: Field name.
+        """
+        return self._snode._name
+
+    def parent(self, n=1):
+        """Gets an ancestor of the representative SNode in the SNode tree.
+
+        Args:
+            n (int): the number of levels going up from the representative SNode.
+
+        Returns:
+            SNode: The n-th parent of the representative SNode.
+        """
+        return self.snode.parent(n)
+
+    def _get_field_members(self):
+        """Gets field members.
+
+        Returns:
+            List[Expr]: Field members.
+        """
+        return self.vars
+
+    def _loop_range(self):
+        """Gets SNode of representative field member for loop range info.
+
+        Returns:
+            taichi_python.SNode: SNode of representative (first) field member.
+        """
+        return self.vars[0].ptr.snode()
+
+    def _set_grad(self, grad):
+        """Sets corresponding grad field (reverse mode).
+        Args:
+            grad (Field): Corresponding grad field.
+        """
+        self.grad = grad
+
+    def _set_dual(self, dual):
+        """Sets corresponding dual field (forward mode).
+
+        Args:
+            dual (Field): Corresponding dual field.
+        """
+        self.dual = dual
+
+    @python_scope
+    def fill(self, val):
+        """Fills `self` with a specific value.
+
+        Args:
+            val (Union[int, float]): Value to fill.
+        """
+        raise NotImplementedError()
+
+    @python_scope
+    def to_numpy(self, dtype=None):
+        """Converts `self` to a numpy array.
+
+        Args:
+            dtype (DataType, optional): The desired data type of returned numpy array.
+
+        Returns:
+            numpy.ndarray: The result numpy array.
+        """
+        raise NotImplementedError()
+
+    @python_scope
+    def to_torch(self, device=None):
+        """Converts `self` to a torch tensor.
+
+        Args:
+            device (torch.device, optional): The desired device of returned tensor.
+
+        Returns:
+            torch.tensor: The result torch tensor.
+        """
+        raise NotImplementedError()
+
+    @python_scope
+    def to_paddle(self, place=None):
+        """Converts `self` to a paddle tensor.
+
+        Args:
+            place (paddle.CPUPlace()/CUDAPlace(n), optional): The desired place of returned tensor.
+
+        Returns:
+            paddle.Tensor: The result paddle tensor.
+        """
+        raise NotImplementedError()
+
+    @python_scope
+    def from_numpy(self, arr):
+        """Loads all elements from a numpy array.
+
+        The shape of the numpy array needs to be the same as `self`.
+
+        Args:
+            arr (numpy.ndarray): The source numpy array.
+        """
+        raise NotImplementedError()
+
+    @python_scope
+    def _from_external_arr(self, arr):
+        raise NotImplementedError()
+
+    @python_scope
+    def from_torch(self, arr):
+        """Loads all elements from a torch tensor.
+
+        The shape of the torch tensor needs to be the same as `self`.
+
+        Args:
+            arr (torch.tensor): The source torch tensor.
+        """
+        self._from_external_arr(arr.contiguous())
+
+    @python_scope
+    def from_paddle(self, arr):
+        """Loads all elements from a paddle tensor.
+
+        The shape of the paddle tensor needs to be the same as `self`.
+
+        Args:
+            arr (paddle.Tensor): The source paddle tensor.
+        """
+        self.from_numpy(arr)
+
+    @python_scope
+    def copy_from(self, other):
+        """Copies all elements from another field.
+
+        The shape of the other field needs to be the same as `self`.
+
+        Args:
+            other (Field): The source field.
+        """
+        if not isinstance(other, Field):
+            raise TypeError("Cannot copy from a non-field object")
+        if self.shape != other.shape:
+            raise ValueError(f"ti.field shape {self.shape} does not match" f" the source field shape {other.shape}")
+        from taichi._kernels import tensor_to_tensor  # pylint: disable=C0415
+
+        tensor_to_tensor(self, other)
+
+    @python_scope
+    def __setitem__(self, key, value):
+        """Sets field element in Python scope.
+
+        Args:
+            key (Union[List[int], int, None]): Coordinates of the field element.
+            value (element type): Value to set.
+        """
+        raise NotImplementedError()
+
+    @python_scope
+    def __getitem__(self, key):
+        """Gets field element in Python scope.
+
+        Args:
+            key (Union[List[int], int, None]): Coordinates of the field element.
+
+        Returns:
+            element type: Value retrieved.
+        """
+        raise NotImplementedError()
+
+    def __str__(self):
+        if taichi.lang.impl.inside_kernel():
+            return self.__repr__()  # make pybind11 happy, see Matrix.__str__
+        if self._snode.ptr is None:
+            return "<Field: Definition of this field is incomplete>"
+        return str(self.to_numpy())
+
+    def _pad_key(self, key):
+        if key is None:
+            key = ()
+        if not isinstance(key, (tuple, list)):
+            key = (key,)
+
+        if len(key) != len(self.shape):
+            raise AssertionError("Slicing is not supported on ti.field")
+
+        return key + ((0,) * (_ti_core.get_max_num_indices() - len(key)))
+
+    def _initialize_host_accessors(self):
+        if self.host_accessors:
+            return
+        taichi.lang.impl.get_runtime().materialize()
+        self.host_accessors = [SNodeHostAccessor(e.ptr.snode()) for e in self.vars]
+
+    def _host_access(self, key):
+        return [SNodeHostAccess(e, key) for e in self.host_accessors]
+
+    def __iter__(self):
+        raise NotImplementedError("Struct for is only available in Taichi scope.")
+
+
+class ScalarField(Field):
+    """Taichi scalar field with SNode implementation.
+
+    Args:
+        var (Expr): Field member.
+    """
+
+    def __init__(self, var):
+        super().__init__([var])
+
+    def fill(self, val):
+        """Fills this scalar field with a specified value."""
+        if in_python_scope():
+            from taichi._kernels import fill_field  # pylint: disable=C0415
+
+            fill_field(self, val)
+        else:
+            from taichi._funcs import field_fill_taichi_scope  # pylint: disable=C0415
+
+            field_fill_taichi_scope(self, val)
+
+    @python_scope
+    def to_numpy(self, dtype=None):
+        """Converts this field to a `numpy.ndarray`."""
+        if self.parent()._snode.ptr.type == _ti_core.SNodeType.dynamic:
+            warn(
+                "You are trying to convert a dynamic snode to a numpy array, be aware that inactive items in the snode will be converted to zeros in the resulting array."
+            )
+        if dtype is None:
+            dtype = to_numpy_type(self.dtype)
+        import numpy as np  # pylint: disable=C0415
+
+        arr = np.zeros(shape=self.shape, dtype=dtype)
+        from taichi._kernels import tensor_to_ext_arr  # pylint: disable=C0415
+
+        tensor_to_ext_arr(self, arr)
+        taichi.lang.runtime_ops.sync()
+        return arr
+
+    @python_scope
+    def to_torch(self, device=None):
+        """Converts this field to a `torch.tensor`."""
+        import torch  # pylint: disable=C0415
+
+        # pylint: disable=E1101
+        arr = torch.zeros(size=self.shape, dtype=to_pytorch_type(self.dtype), device=device)
+        from taichi._kernels import tensor_to_ext_arr  # pylint: disable=C0415
+
+        tensor_to_ext_arr(self, arr)
+        taichi.lang.runtime_ops.sync()
+        return arr
+
+    @python_scope
+    def to_paddle(self, place=None):
+        """Converts this field to a `paddle.Tensor`."""
+        import paddle  # pylint: disable=C0415
+
+        # pylint: disable=E1101
+        # paddle.empty() doesn't support argument `place``
+        arr = paddle.to_tensor(paddle.zeros(self.shape, to_paddle_type(self.dtype)), place=place)
+        from taichi._kernels import tensor_to_ext_arr  # pylint: disable=C0415
+
+        tensor_to_ext_arr(self, arr)
+        taichi.lang.runtime_ops.sync()
+        return arr
+
+    @python_scope
+    def _from_external_arr(self, arr):
+        if len(self.shape) != len(arr.shape):
+            raise ValueError(f"ti.field shape {self.shape} does not match" f" the numpy array shape {arr.shape}")
+        for i, _ in enumerate(self.shape):
+            if self.shape[i] != arr.shape[i]:
+                raise ValueError(f"ti.field shape {self.shape} does not match" f" the numpy array shape {arr.shape}")
+        from taichi._kernels import ext_arr_to_tensor  # pylint: disable=C0415
+
+        ext_arr_to_tensor(arr, self)
+        taichi.lang.runtime_ops.sync()
+
+    @python_scope
+    def from_numpy(self, arr):
+        """Copies the data from a `numpy.ndarray` into this field."""
+        if not arr.flags.c_contiguous:
+            import numpy as np  # pylint: disable=C0415
+
+            arr = np.ascontiguousarray(arr)
+        self._from_external_arr(arr)
+
+    @python_scope
+    def __setitem__(self, key, value):
+        self._initialize_host_accessors()
+        self.host_accessors[0].setter(value, *self._pad_key(key))
+
+    @python_scope
+    def __getitem__(self, key):
+        self._initialize_host_accessors()
+        # Check for potential slicing behaviour
+        # for instance: x[0, :]
+        padded_key = self._pad_key(key)
+        import numpy as np  # pylint: disable=C0415
+
+        for key in padded_key:
+            if not isinstance(key, (int, np.integer)):
+                raise TypeError(
+                    f"Detected illegal element of type: {type(key)}. "
+                    f"Please be aware that slicing a ti.field is not supported so far."
+                )
+        return self.host_accessors[0].getter(*padded_key)
+
+    def __repr__(self):
+        # make interactive shell happy, prevent materialization
+        return "<ti.field>"
+
+
+class SNodeHostAccessor:
+    def __init__(self, snode):
+        if _ti_core.is_real(snode.data_type()):
+            write_func = snode.write_float
+            read_func = snode.read_float
+        else:
+
+            def write_func(key, value):
+                if value >= 0:
+                    snode.write_uint(key, value)
+                else:
+                    snode.write_int(key, value)
+
+            if _ti_core.is_signed(snode.data_type()):
+                read_func = snode.read_int
+            else:
+                read_func = snode.read_uint
+
+        def getter(*key):
+            assert len(key) == _ti_core.get_max_num_indices()
+            return read_func(key)
+
+        def setter(value, *key):
+            assert len(key) == _ti_core.get_max_num_indices()
+            write_func(key, value)
+            # same as above
+            if (
+                impl.get_runtime().target_tape
+                and impl.get_runtime().target_tape.grad_checker
+                and not impl.get_runtime().grad_replaced
+            ):
+                for x in impl.get_runtime().target_tape.grad_checker.to_check:
+                    assert snode != x.snode.ptr, "Overwritten is prohibitive when doing grad check."
+                impl.get_runtime().target_tape.insert(write_func, (key, value))
+
+        self.getter = getter
+        self.setter = setter
+
+
+class SNodeHostAccess:
+    def __init__(self, accessor, key):
+        self.accessor = accessor
+        self.key = key
+
+
+class BitpackedFields:
+    """Taichi bitpacked fields, where fields with quantized types are packed together.
+
+    Args:
+        max_num_bits (int): Maximum number of bits all fields inside can occupy in total. Only 32 or 64 is allowed.
+    """
+
+    def __init__(self, max_num_bits):
+        self.fields = []
+        self.bit_struct_type_builder = _ti_core.BitStructTypeBuilder(max_num_bits)
+
+    def place(self, *args, shared_exponent=False):
+        """Places a list of fields with quantized types inside.
+
+        Args:
+            *args (List[Field]): A list of fields with quantized types to place.
+            shared_exponent (bool): Whether the fields have a shared exponent.
+        """
+        if shared_exponent:
+            self.bit_struct_type_builder.begin_placing_shared_exponent()
+        count = 0
+        for arg in args:
+            assert isinstance(arg, Field)
+            for var in arg._get_field_members():
+                self.fields.append((var.ptr, self.bit_struct_type_builder.add_member(var.ptr.get_dt())))
+                count += 1
+        if shared_exponent:
+            self.bit_struct_type_builder.end_placing_shared_exponent()
+            if count <= 1:
+                raise TaichiSyntaxError("At least 2 fields need to be placed when shared_exponent=True")
+
+
+__all__ = ["BitpackedFields", "Field", "ScalarField"]