sum_dims.reset(sum_dims_raw);
  }

  // Reduce sum the output on all dimensions.
  TF_RETURN_IF_ERROR(ops::Sum(ctx, model_out.get(), sum_dims.get(), &outputs[0],
                              /*keep_dims=*/false, "sum_output"));
  return absl::OkStatus();
}
// ========================= End Helper Functions==============================

absl::Status CalcNumericalGrad(AbstractContext* ctx, Model forward,

  TFE_OpAddInput(op, input, status);
  CHECK_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  TFE_OpAddInput(op, axis, status);
  CHECK_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  TFE_OpSetAttrBool(op, "keep_dims", 1);
  TFE_OpSetAttrType(op, "Tidx", TF_INT32);
  TF_DeleteStatus(status);
  TFE_OpSetAttrType(op, "T", TFE_TensorHandleDataType(input));

  return op;
}

        mean, var = tf.nn.moments(self.kernel, axes=[0, 1, 2], keepdims=True)
        return self.convolution_op(inputs, (self.kernel - mean) / tf.sqrt(var +
        1e-10))` Alternatively, you can override `convolution_op`: `python class
        StandardizedConv2D(tf.keras.Layer): def convolution_op(self, inputs,
        kernel): mean, var = tf.nn.moments(kernel, axes=[0, 1, 2],
        keepdims=True) # Author code uses std + 1e-5 return