* **Machine Learning**: it normally requires lots of "matrix" and "vector" multiplications. Think of a huge spreadsheet with numbers and multiplying all of them together at the same time.

        return new IllegalArgumentException(
            "Multiple entries with same key: " + key + "=" + value1 + " and " + key + "=" + value2);
      }
    }
  }

  /**
   * Returns an immutable map containing the same entries as {@code map}. The returned map iterates
   * over entries in the same order as the {@code entrySet} of the original map. If {@code map}

// Converts tfl.broadcast_to to tfl.reshape if input and output have the same
// number of elements.
// CHECK-LABEL: broadcast_to_to_reshape
func.func @broadcast_to_to_reshape(%arg0: tensor<4x4x4xf32>, %arg1 : tensor<4xi32>) -> tensor<1x4x4x4xf32> {
  %0 = "tfl.broadcast_to"(%arg0, %arg1) : (tensor<4x4x4xf32>, tensor<4xi32>) -> tensor<1x4x4x4xf32>
  // CHECK: "tfl.reshape"
  // CHECK-SAME: (tensor<4x4x4xf32>, tensor<4xi32>) -> tensor<1x4x4x4xf32>

### Create the `Settings` object

Import `BaseSettings` from Pydantic and create a sub-class, very much like with a Pydantic model.

The same way as with Pydantic models, you declare class attributes with type annotations, and possibly default values.

You can use all the same validation features and tools you use for Pydantic models, like different data types and additional validations with `Field()`.

=== "Pydantic v2"

    std::string saved_model_dir = "");

// Creates function pass to cluster TensorFlow ops by host. The program
// generated by this pass will have one function per host where all operations
// in the same function are placed on the same host. Each result of the per-host
// function will have a "tf.device" attribute which specifies the device
// assignment of the result.
std::unique_ptr<OperationPass<mlir::ModuleOp>> CreateClusterTFOpsByHostPass();

// than once for the same mux will result in a panic.
func InstallLivezHandler(mux mux, checks ...HealthChecker) {
	InstallPathHandler(mux, "/livez", checks...)
}

// InstallPathHandler registers handlers for health checking on
// a specific path to mux. *All handlers* for the path must be
// specified in exactly one call to InstallPathHandler. Calling
// InstallPathHandler more than once for the same path and mux will

    value {
      type: DT_INT32
    }
  }
}

# MLIR-LABEL: func @main
# MLIR-SAME:  (%[[ARG_0:[a-z0-9]+]]: tensor<1x1x1x256x!quant.uniform<i8:f32, 0.21632751372549019:27>>) -> tensor<1x6x31x!quant.uniform<i8:f32, 0.09363494573854933:22>>
# MLIR-SAME:  control_outputs = ""
# MLIR-SAME:  inputs = "input"
# MLIR-SAME:  outputs = "output"
# MLIR:         %[[shape:.*]] = arith.constant dense<[1, -1, 31]> : tensor<3xi32>

  }
  // CHECK: func.func @main
  // CHECK: %[[INPUT:.*]]:3 = "tf.XlaCallModule"()
  // CHECK-SAME: _entry_function = @_stablehlo_main_0
  // CHECK: %[[ADD:.*]] = "tf.StatefulPartitionedCall"(%[[INPUT]]#1, %[[INPUT]]#2)
  // CHECK-SAME: f = @some_func
  // CHECK: "tf.PartitionedCall"(%[[ADD]], %[[INPUT]]#0)
  // CHECK-SAME: f = @some_other_func
  // CHECK: return

   * </ol>
   *
   * @param multimap the presumed-immutable multimap
   * @param sampleKey a key of the same type as that contained by {@code multimap}. {@code multimap}
   *     may or may not have {@code sampleKey} as a key.
   * @param sampleValue a key of the same type as that contained by {@code multimap}. {@code
   *     multimap} may or may not have {@code sampleValue} as a key.
   */

  public void testConstructor2NPE() {
    double[] a = null;
    assertThrows(NullPointerException.class, () -> new AtomicDoubleArray(a));
  }

  /** constructor with array is of same size and has all elements */
  public void testConstructor2() {
    AtomicDoubleArray aa = new AtomicDoubleArray(VALUES);
    assertEquals(VALUES.length, aa.length());
    for (int i = 0; i < VALUES.length; i++) {