Add jax implementation of `pt.linalg.pinv` (#294)

pytensor/link/jax/dispatch/nlinalg.py

@@ -3,7 +3,16 @@ import jax.numpy as jnp
 from pytensor.link.jax.dispatch import jax_funcify
 from pytensor.tensor.blas import BatchedDot
 from pytensor.tensor.math import Dot, MaxAndArgmax
-from pytensor.tensor.nlinalg import SVD, Det, Eig, Eigh, MatrixInverse, QRFull, SLogDet
+from pytensor.tensor.nlinalg import (
+    SVD,
+    Det,
+    Eig,
+    Eigh,
+    MatrixInverse,
+    MatrixPinv,
+    QRFull,
+    SLogDet,
+)
 
 
 @jax_funcify.register(SVD)
@@ -77,6 +86,14 @@ def jax_funcify_Dot(op, **kwargs):
     return dot
 
 
+@jax_funcify.register(MatrixPinv)
+def jax_funcify_Pinv(op, **kwargs):
+    def pinv(x):
+        return jnp.linalg.pinv(x)
+
+    return pinv
+
+
 @jax_funcify.register(BatchedDot)
 def jax_funcify_BatchedDot(op, **kwargs):
     def batched_dot(a, b):

tests/link/jax/test_nlinalg.py

@@ -134,3 +134,12 @@ def test_tensor_basics():
     out = at_max(y)
     fgraph = FunctionGraph([y], [out])
     compare_jax_and_py(fgraph, [get_test_value(i) for i in fgraph.inputs])
+
+
+def test_pinv():
+    x = matrix("x")
+    x_inv = at_nlinalg.pinv(x)
+
+    fgraph = FunctionGraph([x], [x_inv])
+    x_np = np.array([[1.0, 2.0], [3.0, 4.0]])
+    compare_jax_and_py(fgraph, [x_np])