keras-nightly 3.12.0.dev2025090303__py3-none-any.whl → 3.12.0.dev2025090503__py3-none-any.whl

keras/_tf_keras/keras/ops/__init__.py

@@ -208,6 +208,7 @@ from keras.src.ops.numpy import isnan as isnan
 from keras.src.ops.numpy import isneginf as isneginf
 from keras.src.ops.numpy import isposinf as isposinf
 from keras.src.ops.numpy import kaiser as kaiser
+from keras.src.ops.numpy import kron as kron
 from keras.src.ops.numpy import left_shift as left_shift
 from keras.src.ops.numpy import less as less
 from keras.src.ops.numpy import less_equal as less_equal

keras/_tf_keras/keras/ops/numpy/__init__.py

@@ -96,6 +96,7 @@ from keras.src.ops.numpy import isnan as isnan
 from keras.src.ops.numpy import isneginf as isneginf
 from keras.src.ops.numpy import isposinf as isposinf
 from keras.src.ops.numpy import kaiser as kaiser
+from keras.src.ops.numpy import kron as kron
 from keras.src.ops.numpy import left_shift as left_shift
 from keras.src.ops.numpy import less as less
 from keras.src.ops.numpy import less_equal as less_equal

keras/ops/__init__.py

@@ -208,6 +208,7 @@ from keras.src.ops.numpy import isnan as isnan
 from keras.src.ops.numpy import isneginf as isneginf
 from keras.src.ops.numpy import isposinf as isposinf
 from keras.src.ops.numpy import kaiser as kaiser
+from keras.src.ops.numpy import kron as kron
 from keras.src.ops.numpy import left_shift as left_shift
 from keras.src.ops.numpy import less as less
 from keras.src.ops.numpy import less_equal as less_equal

keras/ops/numpy/__init__.py

@@ -96,6 +96,7 @@ from keras.src.ops.numpy import isnan as isnan
 from keras.src.ops.numpy import isneginf as isneginf
 from keras.src.ops.numpy import isposinf as isposinf
 from keras.src.ops.numpy import kaiser as kaiser
+from keras.src.ops.numpy import kron as kron
 from keras.src.ops.numpy import left_shift as left_shift
 from keras.src.ops.numpy import less as less
 from keras.src.ops.numpy import less_equal as less_equal

keras/src/backend/jax/numpy.py

@@ -809,6 +809,12 @@ def isposinf(x):
     return jnp.isposinf(x)
 
 
+def kron(x1, x2):
+    x1 = convert_to_tensor(x1)
+    x2 = convert_to_tensor(x2)
+    return jnp.kron(x1, x2)
+
+
 def less(x1, x2):
     x1 = convert_to_tensor(x1)
     x2 = convert_to_tensor(x2)

keras/src/backend/numpy/numpy.py

@@ -742,6 +742,13 @@ def isposinf(x):
     return np.isposinf(x)
 
 
+def kron(x1, x2):
+    x1 = convert_to_tensor(x1)
+    x2 = convert_to_tensor(x2)
+    dtype = dtypes.result_type(x1.dtype, x2.dtype)
+    return np.kron(x1, x2).astype(dtype)
+
+
 def less(x1, x2):
     return np.less(x1, x2)
 

keras/src/backend/openvino/numpy.py

@@ -978,6 +978,10 @@ def isposinf(x):
     )
 
 
+def kron(x1, x2):
+    raise NotImplementedError("`kron` is not supported with openvino backend")
+
+
 def less(x1, x2):
     element_type = None
     if isinstance(x1, OpenVINOKerasTensor):

keras/src/backend/tensorflow/numpy.py

@@ -1711,6 +1711,52 @@ def isposinf(x):
     return tf.math.equal(x, tf.constant(float("inf"), dtype=x.dtype))
 
 
+def kron(x1, x2):
+    x1 = convert_to_tensor(x1)
+    x2 = convert_to_tensor(x2)
+
+    dtype = dtypes.result_type(x1.dtype, x2.dtype)
+    x1 = tf.cast(x1, dtype)
+    x2 = tf.cast(x2, dtype)
+
+    ndim_x1 = tf.rank(x1)
+    ndim_x2 = tf.rank(x2)
+
+    def expand_front(x, num):
+        for _ in range(num):
+            x = tf.expand_dims(x, axis=0)
+        return x
+
+    x1 = tf.cond(
+        ndim_x1 < ndim_x2,
+        lambda: expand_front(x1, ndim_x2 - ndim_x1),
+        lambda: x1,
+    )
+    x2 = tf.cond(
+        ndim_x2 < ndim_x1,
+        lambda: expand_front(x2, ndim_x1 - ndim_x2),
+        lambda: x2,
+    )
+
+    x1_reshaped = tf.reshape(
+        x1,
+        tf.reshape(
+            tf.stack([tf.shape(x1), tf.ones_like(tf.shape(x1))], axis=1), [-1]
+        ),
+    )
+    x2_reshaped = tf.reshape(
+        x2,
+        tf.reshape(
+            tf.stack([tf.ones_like(tf.shape(x2)), tf.shape(x2)], axis=1), [-1]
+        ),
+    )
+
+    out = tf.multiply(x1_reshaped, x2_reshaped)
+    out_shape = tf.multiply(tf.shape(x1), tf.shape(x2))
+    out = tf.reshape(out, out_shape)
+    return out
+
+
 def less(x1, x2):
     x1 = convert_to_tensor(x1)
     x2 = convert_to_tensor(x2)

keras/src/backend/torch/numpy.py

@@ -945,6 +945,12 @@ def isposinf(x):
     return torch.isposinf(x)
 
 
+def kron(x1, x2):
+    x1 = convert_to_tensor(x1)
+    x2 = convert_to_tensor(x2)
+    return torch.kron(x1, x2)
+
+
 def less(x1, x2):
     x1, x2 = convert_to_tensor(x1), convert_to_tensor(x2)
     return torch.less(x1, x2)

keras/src/models/model.py

@@ -9,6 +9,7 @@ from keras.src.api_export import keras_export
 from keras.src.layers.layer import Layer
 from keras.src.models.variable_mapping import map_saveable_variables
 from keras.src.quantizers.gptq_config import GPTQConfig
+from keras.src.quantizers.gptq_core import gptq_quantize
 from keras.src.saving import saving_api
 from keras.src.trainers import trainer as base_trainer
 from keras.src.utils import summary_utils
@@ -440,8 +441,7 @@ class Model(Trainer, base_trainer.Trainer, Layer):
                     "The `config` argument must be of type "
                     "`keras.quantizers.GPTQConfig`."
                 )
-            # The config object's own quantize method drives the process
-            config.quantize(self)
+            gptq_quantize(self, config)
             return
 
         # For all other modes, verify that a config object was not passed.

keras/src/ops/numpy.py

@@ -3845,6 +3845,49 @@ def isposinf(x):
     return backend.numpy.isposinf(x)
 
 
+class Kron(Operation):
+    def call(self, x1, x2):
+        return backend.numpy.kron(x1, x2)
+
+    def compute_output_spec(self, x1, x2):
+        x1_shape = getattr(x1, "shape", [])
+        x2_shape = getattr(x2, "shape", [])
+
+        def _mul_shape_dim(a, b):
+            if a is None or b is None:
+                return None
+            return a * b
+
+        output_shape = tuple(
+            _mul_shape_dim(a, b) for a, b in zip(x1_shape, x2_shape)
+        )
+
+        x1_type = backend.standardize_dtype(getattr(x1, "dtype", type(x1)))
+        x2_type = backend.standardize_dtype(getattr(x2, "dtype", type(x2)))
+        dtype = dtypes.result_type(x1_type, x2_type)
+        return KerasTensor(output_shape, dtype=dtype)
+
+
+@keras_export(["keras.ops.kron", "keras.ops.numpy.kron"])
+def kron(x1, x2):
+    """Kronecker product of `x1` and `x2`.
+
+    Computes the Kronecker product of two input tensors. If `x1` has shape
+    `(a0, a1, ..., an)` and `x2` has shape `(b0, b1, ..., bn)`, then the
+    output will have shape `(a0*b0, a1*b1, ..., an*bn)`.
+
+    Args:
+        x1: First input tensor.
+        x2: Second input tensor.
+
+    Returns:
+        A tensor representing the Kronecker product of `x1` and `x2`.
+    """
+    if any_symbolic_tensors((x1, x2)):
+        return Kron().symbolic_call(x1, x2)
+    return backend.numpy.kron(x1, x2)
+
+
 class Less(Operation):
     def call(self, x1, x2):
         return backend.numpy.less(x1, x2)