Make repeated indexes work

pytensor/tensor/einsum.py

@@ -8,7 +8,6 @@
 from pytensor.compile.builders import OpFromGraph
 from pytensor.tensor.basic import (
     arange,
-    expand_dims,
     get_vector_length,
     stack,
     transpose,
@@ -36,9 +35,10 @@ def __init__(
 
 
 def _iota(shape: TensorVariable, axis: int) -> TensorVariable:
-    axis = normalize_axis_index(axis, get_vector_length(shape))
+    len_shape = get_vector_length(shape)
+    axis = normalize_axis_index(axis, len_shape)
     values = arange(shape[axis])
-    return broadcast_to(shape_padright(values, axis), shape)
+    return broadcast_to(shape_padright(values, len_shape - axis - 1), shape)
 
 
 def _delta(shape, axes: Sequence[int]) -> TensorVariable:
@@ -47,7 +47,7 @@ def _delta(shape, axes: Sequence[int]) -> TensorVariable:
     iotas = [_iota(base_shape, i) for i in range(len(axes))]
     eyes = [eq(i1, i2) for i1, i2 in pairwise(iotas)]
     result = reduce(and_, eyes)
-    return broadcast_to(expand_dims(result, tuple(axes)), shape)
+    return broadcast_to(result, shape)
 
 
 def _removechars(s, chars):

tests/tensor/test_einsum.py

@@ -4,6 +4,39 @@
 import pytest
 
 import pytensor.tensor as pt
+from pytensor import Mode
+from pytensor.tensor.einsum import _delta, _iota
+
+
+def test_iota():
+    mode = Mode(linker="py", optimizer=None)
+    np.testing.assert_allclose(
+        _iota((4, 8), 0).eval(mode=mode),
+        [
+            [0, 0, 0, 0, 0, 0, 0, 0],
+            [1, 1, 1, 1, 1, 1, 1, 1],
+            [2, 2, 2, 2, 2, 2, 2, 2],
+            [3, 3, 3, 3, 3, 3, 3, 3],
+        ],
+    )
+
+    np.testing.assert_allclose(
+        _iota((4, 8), 1).eval(mode=mode),
+        [
+            [0, 1, 2, 3, 4, 5, 6, 7],
+            [0, 1, 2, 3, 4, 5, 6, 7],
+            [0, 1, 2, 3, 4, 5, 6, 7],
+            [0, 1, 2, 3, 4, 5, 6, 7],
+        ],
+    )
+
+
+def test_delta():
+    mode = Mode(linker="py", optimizer=None)
+    np.testing.assert_allclose(
+        _delta((2, 2), (0, 1)).eval(mode=mode),
+        [[1.0, 0.0], [0.0, 1.0]],
+    )
 
 
 @pytest.mark.parametrize(