feat(frontend): Added logsumexp function in paddlepaddle frontend

ivy/functional/frontends/paddle/math.py

@@ -417,6 +417,29 @@ def logit(x, eps=None, name=None):
     return ivy.logit(x, eps=eps)
 
 
+@with_supported_dtypes({"2.5.1 and below": ("float32", "float64")}, "paddle")
+@to_ivy_arrays_and_back
+def logsumexp(x, axis=None, y=None):
+    x = ivy.asarray(x)
+    if y is not None:
+        y = ivy.asarray(y)
+        x = ivy.where(y != 0, x, -ivy.inf)
+    if axis is None:
+        amax = ivy.max(x)
+        expsub = ivy.exp(x - amax)
+        sumexp = ivy.sum(expsub)
+        out = ivy.log(sumexp) + amax
+    else:
+        amax = ivy.max(x, axis=axis, keepdims=True)
+        expsub = ivy.exp(x - amax)
+        sumexp = ivy.sum(expsub, axis=axis, keepdims=True)
+        out = ivy.log(sumexp) + amax
+    if y is not None:
+        sign = ivy.stop_gradient(ivy.sign(sumexp))
+        out = ivy.where(sign < 0, ivy.nan, out)
+    return out
+
+
 @with_supported_dtypes(
     {"2.5.2 and below": ("float32", "float64", "int32", "int64")}, "paddle"
 )

ivy_tests/test_ivy/test_frontends/test_paddle/test_math.py

@@ -1720,6 +1720,39 @@ def test_paddle_logit(
     )
 
 
+# logsumexp
+@handle_frontend_test(
+    fn_tree="paddle.tensor.math.logsumexp",
+    dtype_and_x=helpers.dtype_and_values(
+        available_dtypes=helpers.get_dtypes("valid"),
+        max_num_dims=4,
+        num_arrays=2,
+        allow_inf=False,
+        shared_dtype=True,
+    ),
+)
+def test_paddle_logsumexp(
+    *,
+    dtype_and_x,
+    on_device,
+    fn_tree,
+    backend_fw,
+    frontend,
+    test_flags,
+):
+    input_dtypes, xs = dtype_and_x
+    helpers.test_frontend_function(
+        input_dtypes=input_dtypes,
+        backend_to_test=backend_fw,
+        frontend=frontend,
+        test_flags=test_flags,
+        fn_tree=fn_tree,
+        on_device=on_device,
+        x=xs[0],
+        axis=None,
+    )
+
+
 # max
 @handle_frontend_test(
     fn_tree="paddle.max",