Remove dependencies

keras_nlp/src/tests/test_case.py

@@ -77,7 +77,7 @@ def assertAllEqual(self, x1, x2, msg=None):
         super().assertAllEqual(x1, x2, msg=msg)
 
     def assertDTypeEqual(self, x, expected_dtype, msg=None):
-        input_dtype = keras.utils.standardize_dtype(x.dtype)
+        input_dtype = keras.backend.standardize_dtype(x.dtype)
         super().assertEqual(input_dtype, expected_dtype, msg=msg)
 
     def run_layer_test(
@@ -127,8 +127,8 @@ def run_output_asserts(layer, output, eager=False):
             )
             output_dtype = tree.flatten(output)[0].dtype
             self.assertEqual(
-                keras.utils.standardize_dtype(layer.dtype),
-                keras.utils.standardize_dtype(output_dtype),
+                keras.backend.standardize_dtype(layer.dtype),
+                keras.backend.standardize_dtype(output_dtype),
                 msg="Unexpected output dtype",
             )
             if eager and expected_output_data is not None:

keras_nlp/src/utils/tensor_utils.py

@@ -71,7 +71,7 @@ def convert_to_backend_tensor_or_python_list(x):
     if isinstance(x, tf.RaggedTensor) or getattr(x, "dtype", None) == tf.string:
         return tensor_to_list(x)
     dtype = getattr(x, "dtype", "float32")
-    dtype = keras.utils.standardize_dtype(dtype)
+    dtype = keras.backend.standardize_dtype(dtype)
     return ops.convert_to_tensor(x, dtype=dtype)
 
 
@@ -160,15 +160,15 @@ def is_tensor_type(x):
 
 
 def is_float_dtype(dtype):
-    return "float" in keras.utils.standardize_dtype(dtype)
+    return "float" in keras.backend.standardize_dtype(dtype)
 
 
 def is_int_dtype(dtype):
-    return "int" in keras.utils.standardize_dtype(dtype)
+    return "int" in keras.backend.standardize_dtype(dtype)
 
 
 def is_string_dtype(dtype):
-    return "string" in keras.utils.standardize_dtype(dtype)
+    return "string" in keras.backend.standardize_dtype(dtype)
 
 
 def any_equal(inputs, values, padding_mask):

setup.py

@@ -55,13 +55,11 @@ def get_version(rel_path):
     author_email="keras-nlp@google.com",
     license="Apache License 2.0",
     install_requires=[
-        "keras-core",
         "absl-py",
         "numpy",
         "packaging",
         "regex",
         "rich",
-        "dm-tree",
         "kagglehub",
         # Don't require tensorflow-text on MacOS, there are no binaries for ARM.
         # Also, we rely on tensorflow *transitively* through tensorflow-text.