Support numpy.prod operation

keras/src/backend/openvino/excluded_concrete_tests.txt

@@ -41,7 +41,6 @@ NumpyDtypeTest::test_multiply
 NumpyDtypeTest::test_nan
 NumpyDtypeTest::test_outer_
 NumpyDtypeTest::test_power
-NumpyDtypeTest::test_prod
 NumpyDtypeTest::test_quantile
 NumpyDtypeTest::test_ravel
 NumpyDtypeTest::test_repeat
@@ -104,7 +103,6 @@ NumpyOneInputOpsCorrectnessTest::test_pad_int16_constant_2
 NumpyOneInputOpsCorrectnessTest::test_pad_int8_constant_2
 NumpyOneInputOpsCorrectnessTest::test_pad_uint8_constant_2
 NumpyOneInputOpsCorrectnessTest::test_pad_int32_constant_2
-NumpyOneInputOpsCorrectnessTest::test_prod
 NumpyOneInputOpsCorrectnessTest::test_ravel
 NumpyOneInputOpsCorrectnessTest::test_real
 NumpyOneInputOpsCorrectnessTest::test_reciprocal

keras/src/backend/openvino/numpy.py

@@ -1117,8 +1117,67 @@ def pad(x, pad_width, mode="constant", constant_values=None):
     )
 
 
+"""
+Helper Function to convert the string dtype to ov type
+"""
+
+
+def string_to_ov_type(dtype_str):
+    from openvino.runtime import Type
+
+    mapping = {
+        "bool": Type.boolean,
+        "int8": Type.i8,
+        "int16": Type.i16,
+        "int32": Type.i32,
+        "int64": Type.i64,
+        "uint8": Type.u8,
+        "uint16": Type.u16,
+        "uint32": Type.u32,
+        "uint64": Type.u64,
+        "float16": Type.f16,
+        "float32": Type.f32,
+        "float64": Type.f64,
+    }
+    return mapping[dtype_str]
+
+
 def prod(x, axis=None, keepdims=False, dtype=None):
-    raise NotImplementedError("`prod` is not supported with openvino backend")
+    if axis == () or axis == []:
+        return x
+
+    x = get_ov_output(x)
+    x_type = x.get_element_type()
+
+    # Promote dtype if not explicitly specified
+    if dtype is None:
+        if x_type == Type.boolean:
+            promoted_dtype = Type.i32
+        elif x_type in (Type.i8, Type.i16):
+            promoted_dtype = Type.i32
+        elif x_type in (Type.u8, Type.u16):
+            promoted_dtype = Type.u32
+        else:
+            promoted_dtype = x_type
+    else:
+        promoted_dtype = string_to_ov_type(dtype)
+
+    # Cast to promoted dtype if necessary
+    if x_type != promoted_dtype:
+        x = ov_opset.convert(x, promoted_dtype).output(0)
+
+    if axis is None:
+        flatten_shape = ov_opset.constant([-1], Type.i32).output(0)
+        x = ov_opset.reshape(x, flatten_shape, False).output(0)
+        axis = 0
+
+    if isinstance(axis, tuple):
+        axis = list(axis)
+    axis = ov_opset.constant(axis, Type.i32).output(0)
+
+    return OpenVINOKerasTensor(
+        ov_opset.reduce_prod(x, axis, keepdims).output(0)
+    )
 
 
 def quantile(x, q, axis=None, method="linear", keepdims=False):

requirements.txt

@@ -13,6 +13,7 @@ torch-xla==2.6.0;sys_platform != 'darwin'
 # Jax.
 # Pinned to 0.5.0 on CPU. JAX 0.5.1 requires Tensorflow 2.19 for saved_model_test.
 # Note that we test against the latest JAX on GPU.
+
 jax[cpu]==0.5.0
 flax