Remove Matmul Operator in favor of Blockwise Dot

Author: ricardoV94

pytensor/tensor/math.py

@@ -25,11 +25,11 @@
     stack,
     switch,
 )
+from pytensor.tensor.blockwise import Blockwise
 from pytensor.tensor.elemwise import CAReduce, DimShuffle, Elemwise, scalar_elemwise
 from pytensor.tensor.shape import shape, specify_broadcastable
 from pytensor.tensor.type import (
     DenseTensorType,
-    TensorType,
     complex_dtypes,
     continuous_dtypes,
     discrete_dtypes,
@@ -2868,93 +2868,7 @@ def logsumexp(x, axis=None, keepdims=False):
     return log(sum(exp(x), axis=axis, keepdims=keepdims))
 
 
-class MatMul(Op):
-    __props__ = ("dtype",)
-
-    def __init__(self, dtype=None):
-        self.dtype = dtype
-
-    @classmethod
-    def _get_output_shape(cls, x1, x2, shapes, validate=False):
-        x1_shape, x2_shape = shapes
-
-        if x1.ndim == 1 and x2.ndim == 1:
-            if validate and x1_shape[0] != x2_shape[0]:
-                raise ValueError("1d inputs must have the same length.")
-            return ()
-        elif x1.ndim == 1 and x2.ndim > 1:
-            if validate and x1_shape[0] != x2_shape[-2]:
-                raise ValueError(
-                    "length of input 1 must be equal the length "
-                    "of the 2nd-last dimension of input 2"
-                )
-            return x2_shape[:-2] + x2_shape[-1:]
-        elif x1.ndim > 1 and x2.ndim == 1:
-            if validate and x1_shape[-1] != x2_shape[0]:
-                raise ValueError(
-                    "length of input 2 must be equal the length "
-                    "of the last dimension of input 1"
-                )
-            return x1_shape[:-1]
-        elif x1.ndim == 2 and x2.ndim == 2:
-            if validate and x1_shape[-1] != x2_shape[0]:
-                raise ValueError(
-                    "number of columns of input 1 must be equal to "
-                    "the number of rows of input 2"
-                )
-            return x1_shape[:-1] + x2_shape[-1:]
-        elif x1.ndim > 2 and x2.ndim == 2:
-            if validate and x1_shape[-1] != x2_shape[0]:
-                raise ValueError(
-                    "number of rows of input 2 must be equal to "
-                    "the length of the last dimension of input 1"
-                )
-            return x1_shape[:-2] + x1_shape[-2:-1] + x2_shape[-1:]
-        elif x1.ndim == 2 and x2.ndim > 2:
-            if validate and x1_shape[-1] != x2_shape[-2]:
-                raise ValueError(
-                    "number of columns of input 1 must be equal "
-                    "the length of the 2nd-last dimension of input 2"
-                )
-            return x2_shape[:-2] + x1_shape[-2:-1] + x2_shape[-1:]
-        else:
-            if validate:
-                from pytensor.tensor.random.basic import broadcast_shapes
-
-                bshape = broadcast_shapes(x1_shape[:-2], x2_shape[:-2])
-                if x1_shape[-1] != x2_shape[-2]:
-                    raise ValueError(
-                        "length of the last dimension of input 1 must be equal "
-                        "to the length of the 2nd-last dimension of input 2"
-                    )
-            else:
-                from pytensor.tensor.extra_ops import broadcast_shape
-
-                bshape = broadcast_shape(
-                    x1_shape[:-2], x2_shape[:-2], arrays_are_shapes=True
-                )
-            return bshape + x1_shape[-2:-1] + x2_shape[-1:]
-
-    def make_node(self, a, b):
-        a = as_tensor_variable(a)
-        b = as_tensor_variable(b)
-
-        if 0 in {a.ndim, b.ndim}:
-            raise ValueError("inputs to `matmul` cannot be scalar.")
-
-        out_shape = self._get_output_shape(
-            a, b, (a.type.shape, b.type.shape), validate=True
-        )
-        out = TensorType(dtype=self.dtype, shape=out_shape)()
-        return Apply(self, [a, b], [out])
-
-    def perform(self, node, inputs, outputs):
-        x1, x2 = inputs
-        outputs[0][0] = np.matmul(x1, x2, dtype=self.dtype)
-
-    def infer_shape(self, fgraph, node, shapes):
-        x1, x2 = node.inputs
-        return [self._get_output_shape(x1, x2, shapes)]
+_matrix_matrix_matmul = Blockwise(_dot, signature="(n,k),(k,m)->(n,m)")
 
 
 def matmul(x1: "ArrayLike", x2: "ArrayLike", dtype: Optional["DTypeLike"] = None):
@@ -2999,7 +2913,23 @@ def matmul(x1: "ArrayLike", x2: "ArrayLike", dtype: Optional["DTypeLike"] = None
     - Stacks of matrices are broadcast together as if the matrices were elements,
         respecting the signature ``(n, k), (k, m) -> (n, m)``:
     """
-    return MatMul(dtype=dtype)(x1, x2)
+    x1 = as_tensor_variable(x1)
+    x2 = as_tensor_variable(x2)
+    if x1.type.ndim == 0 or x2.type.ndim == 0:
+        raise ValueError("matmul operand cannot be scalar")
+    if x1.type.ndim == 1 and x2.type.ndim == 1:
+        out = _dot(x1, x2)
+    elif x1.type.ndim == 1:
+        out = _matrix_matrix_matmul(x1[None], x2).squeeze(-2)
+    elif x2.type.ndim == 1:
+        out = _matrix_matrix_matmul(x1, x2[:, None]).squeeze(-1)
+    else:
+        out = _matrix_matrix_matmul(x1, x2)
+
+    if dtype is not None:
+        out = out.astype(dtype)
+
+    return out
 
 
 __all__ = [

pytensor/tensor/rewriting/blockwise.py

@@ -3,6 +3,8 @@
 from pytensor.graph.replace import vectorize_node
 from pytensor.graph.rewriting.basic import copy_stack_trace, out2in
 from pytensor.tensor.blockwise import Blockwise
+from pytensor.tensor.math import _matrix_matrix_matmul
+from pytensor.tensor.rewriting.basic import register_canonicalize
 
 
 @node_rewriter([Blockwise])
@@ -40,3 +42,10 @@ def local_useless_unbatched_blockwise(fgraph, node):
     "blockwise",
     position=49,
 )
+
+
+# Avoid redundant cases early on for Ops whose default form is not Blockwised
+@register_canonicalize
+@node_rewriter(tracks=[_matrix_matrix_matmul])
+def local_eager_useless_unbatched_blockwise(fgraph, node):
+    return local_useless_unbatched_blockwise.fn(fgraph, node)

pytensor/tensor/variable.py

@@ -647,8 +647,12 @@ def __rdot__(right, left):
         return at.math.dense_dot(left, right)
 
     dot = __dot__
-    __matmul__ = __dot__
-    __rmatmul__ = __rdot__
+
+    def __matmul__(left, right):
+        return at.math.matmul(left, right)
+
+    def __rmatmul__(right, left):
+        return at.math.matmul(right, left)
 
     def sum(self, axis=None, dtype=None, keepdims=False, acc_dtype=None):
         """See :func:`pytensor.tensor.math.sum`."""
@@ -797,15 +801,15 @@ def choose(self, choices, mode="raise"):
         """
         return at.basic.choose(self, choices, mode="raise")
 
-    def squeeze(self):
+    def squeeze(self, axis=None):
         """
         Remove broadcastable dimensions from the shape of an array.
 
         It returns the input array, but with the broadcastable dimensions
         removed. This is always `x` itself or a view into `x`.
 
         """
-        return at.extra_ops.squeeze(self)
+        return at.extra_ops.squeeze(self, axis=axis)
 
     def compress(self, a, axis=None):
         """Return selected slices only."""

tests/tensor/test_math.py

@@ -30,11 +30,11 @@
     get_underlying_scalar_constant_value,
     switch,
 )
+from pytensor.tensor.blas import Dot22
 from pytensor.tensor.elemwise import CAReduce, Elemwise
 from pytensor.tensor.math import (
     Argmax,
     Dot,
-    MatMul,
     MaxAndArgmax,
     Mean,
     Prod,
@@ -3412,12 +3412,10 @@ def test_log1mexp_grad_lim():
     assert grad_x_fn(-1e-308) != -np.inf
 
 
-class TestMatMul(utt.InferShapeTester):
+class TestMatMul:
     def setup_method(self):
-        super().setup_method()
         self.rng = np.random.default_rng(utt.fetch_seed())
         self.op = matmul
-        self.op_class = MatMul
 
     def _validate_output(self, a, b):
         pytensor_sol = self.op(a, b).eval()
@@ -3467,85 +3465,8 @@ def test_dtype_param(self, dtype):
         sol = self.op([1, 2, 3], [3, 2, 1], dtype=dtype)
         assert sol.eval().dtype == dtype
 
-    @pytest.mark.parametrize(
-        "x1_shape,x2_shape,exp_res,error_regex",
-        [
-            ((1,), (3,), None, "inputs must have the same length"),
-            ((2,), (3, 1), None, "length of input 1.*2nd-last dimension of input 2"),
-            ((2, 5), (3,), None, "length of input 2.*of the last dimension of input 1"),
-            (
-                (2, 5),
-                (3, 4),
-                None,
-                "number of columns of input 1 .* number of rows of input 2",
-            ),
-            (
-                (2, 1, 3),
-                (5, 4),
-                None,
-                "number of rows of input 2 .* last dimension of input 1",
-            ),
-            (
-                (2, 5),
-                (2, 4, 3),
-                None,
-                "number of columns of input 1 .* 2nd-last dimension of input 2",
-            ),
-            (
-                (3, 2, 4, 5),
-                (1, 6, 7),
-                None,
-                "length of the last dimension of input 1 .* 2nd-last dimension of input 2",
-            ),
-            (
-                (4, 5, 4),
-                (3, 2, 2),
-                None,
-                "cannot be broadcast to a single shape",
-            ),
-            (
-                (4, None, 2),
-                (4, 2, None),
-                (4, None, None),
-                None,
-            ),
-        ],
-    )
-    def test_get_output_shape(self, x1_shape, x2_shape, exp_res, error_regex):
-        x1 = tensor(dtype=np.float64, shape=x1_shape)
-        x2 = tensor(dtype=np.float64, shape=x2_shape)
-
-        if error_regex is not None:
-            with pytest.raises(ValueError, match=error_regex):
-                self.op_class._get_output_shape(
-                    x1, x2, (x1_shape, x2_shape), validate=True
-                )
-        else:
-            assert (
-                self.op_class._get_output_shape(
-                    x1, x2, (x1_shape, x2_shape), validate=True
-                )
-                == exp_res
-            )
-
-    def test_infer_shape(self):
-        for shape_x1, shape_x2 in [
-            ((5,), (5,)),
-            ((5,), (2, 5, 3)),
-            ((2, 5, 3), (3,)),
-            ((2, 5), (5, 4)),
-            ((2, 5), (2, 5, 3)),
-            ((2, 1, 3), (3, 4)),
-            ((3, 2, 4, 5), (1, 5, 7)),
-        ]:
-            a = tensor(dtype=config.floatX, shape=shape_x1)
-            b = tensor(dtype=config.floatX, shape=shape_x2)
-            x1 = self.rng.random(shape_x1).astype(config.floatX)
-            x2 = self.rng.random(shape_x2).astype(config.floatX)
-
-            self._compile_and_check(
-                [a, b],
-                [self.op(a, b)],
-                [x1, x2],
-                self.op_class,
-            )
+    def test_dot22_opt(self):
+        x, y = matrices("xy")
+        fn = function([x, y], x @ y, mode="FAST_RUN")
+        [node] = fn.maker.fgraph.apply_nodes
+        assert isinstance(node.op, Dot22)

tests/tensor/test_variable.py

@@ -10,9 +10,9 @@
 from pytensor.compile.mode import get_default_mode
 from pytensor.graph.basic import Constant, equal_computations
 from pytensor.tensor import get_vector_length
-from pytensor.tensor.basic import constant
+from pytensor.tensor.basic import as_tensor, constant
 from pytensor.tensor.elemwise import DimShuffle
-from pytensor.tensor.math import dot, eq
+from pytensor.tensor.math import dot, eq, matmul
 from pytensor.tensor.shape import Shape
 from pytensor.tensor.subtensor import AdvancedSubtensor, Subtensor
 from pytensor.tensor.type import (
@@ -79,16 +79,30 @@ def test_infix_dot_method():
     X = dmatrix("X")
     y = dvector("y")
 
-    res = X @ y
-    exp_res = X.dot(y)
+    res = X.dot(y)
+    exp_res = dot(X, y)
     assert equal_computations([res], [exp_res])
 
     X_val = np.arange(2 * 3).reshape((2, 3))
-    res = X_val @ y
+    res = as_tensor(X_val).dot(y)
     exp_res = dot(X_val, y)
     assert equal_computations([res], [exp_res])
 
 
+def test_infix_matmul_method():
+    X = dmatrix("X")
+    y = dvector("y")
+
+    res = X @ y
+    exp_res = matmul(X, y)
+    assert equal_computations([res], [exp_res])
+
+    X_val = np.arange(2 * 3).reshape((2, 3))
+    res = as_tensor(X_val) @ y
+    exp_res = matmul(X_val, y)
+    assert equal_computations([res], [exp_res])
+
+
 def test_empty_list_indexing():
     ynp = np.zeros((2, 2))[:, []]
     znp = np.zeros((2, 2))[:, ()]