formatting

keras/src/backend/mlx/linalg.py

@@ -7,7 +7,7 @@
 
 
 def det(a):
-    # TODO: Swap to mlx.linalg.det when it's implemented
+    # TODO: Swap to mlx.linalg.det when supported
     a = jnp.array(a)
     output = jnp.linalg.det(a)
     return mx.array(output)

keras/src/backend/mlx/numpy.py

@@ -420,7 +420,7 @@ def dot(x, y):
     x = x.reshape(x.shape[:-1] + (x.shape[-1],) + (1,) * (ndimy - 2))
     r = x @ y
     return r
-    
+
 
 def empty(shape, dtype=None):
     dtype = to_mlx_dtype(dtype or config.floatx())
@@ -856,6 +856,7 @@ def quantile(x, q, axis=None, method="linear", keepdims=False):
     result = mx.array(result)
     return result
 
+
 def ravel(x):
     x = convert_to_tensor(x)
     return x.reshape(-1)
@@ -884,7 +885,8 @@ def repeat(x, repeats, axis=None):
 
     if repeats.size != x.shape[axis]:
         raise ValueError(
-            f"repeats must have same length as axis: got {repeats.size} vs {x.shape[axis]}"
+            "repeats must have same length as axis: "
+            f"got {repeats.size} vs {x.shape[axis]}"
         )
 
     indices = mx.concatenate([mx.full(r, i) for i, r in enumerate(repeats)])
@@ -982,8 +984,7 @@ def tensordot(x1, x2, axes=2):
         return mx.tensordot(x1, x2, axes)
 
     raise ValueError(
-        "`axes` must be an integer or sequence "
-        f"Received: axes={axes}"
+        "`axes` must be an integer or sequence " f"Received: axes={axes}"
     )
 
 
@@ -1196,7 +1197,7 @@ def select(condlist, choicelist, default=0):
 
 
 def slogdet(x):
-    # TODO: Swap to mlx.linalg.slogdet when supported (or use LU factorization and determinant)
+    # TODO: Swap to mlx.linalg.slogdet when supported (or with determinant)
     x = convert_to_tensor(x)
     x = jnp.array(x)
     output = jnp.linalg.slogdet(x)