return concatNoDefensiveCopy(Arrays.copyOf(inputs, inputs.length));
  }

  /**
   * Combines multiple iterators into a single iterator. The returned iterator iterates across the
   * elements of each iterator in {@code inputs}. The input iterators are not polled until
   * necessary.
   *

//
// A register phi must have the phi and all of its inputs allocated to the
// same register. Register phis are spilled similarly to regular ops.
//
// A stack phi must have the phi and all of its inputs allocated to the same
// stack location. Stack phis start out life already spilled - each phi
// input must be a store (using StoreReg) at the end of the corresponding
// predecessor block.

                    def input = inputArtifact.get().asFile
                    if (input.exists()) {
                        println "processing \${input.name}"
                    } else {
                        println "processing missing \${input.name}"
                    }
                    def output = outputs.file(input.name + ".green")
                    output.text = input.list().length + ".green"
                }

    if (!filter_type || filter_type.getRank() != 4 ||
        !filter_type.hasStaticShape())
      return failure();

    Value input = tf_op.getInput();
    RankedTensorType input_type =
        mlir::dyn_cast<RankedTensorType>(input.getType());
    // Only rank size four input will be only available by the tf.Conv2D
    // operator verification.
    if (!input_type || input_type.isDynamicDim(3)) {
      return failure();

        ClosingFuture.whenAllComplete(ImmutableList.of(input1, input2Failed))
            .call(
                new CombiningCallable<TestCloseable>() {
                  @Override
                  public TestCloseable call(DeferredCloser closer, Peeker peeker) throws Exception {
                    closer.eventuallyClose(closeable1, closingExecutor);
                    assertThat(peeker.getDone(input1)).isSameInstanceAs("value1");

                void transform(TransformOutputs outputs) {
                    def input = inputArtifact.get().asFile
                    println "converting \${input.name} to red"
                    assert input.file
                    def output = outputs.file(input.name + ".red")
                    output.text = input.text + ".red"
                }
            }

            dependencies {

// Predicate denote logical formulas and mapping a node `n` to a predicate
// `pred` implies that `n` is live whenever `pred` is true.  Then we can deduce
// mismatching liveness in the inputs to node by comparing the predicate those
// inputs are mapped to.  The core logic of this pass resides in creating the
// map from TensorFlow nodes to predicates.
//
//
// MAPPING NODES TO PREDICATES, MODULO CYCLES

                            &output_permutation, &call_node_def_));

    // Apply the input/output permutations to the 'args_by_...' and 'results_'
    // mappings, so when we build edges in BuildOutputGraph() we
    // connect them to the right input/output positions.
    if (input_permutation.size() != args_by_src_.size()) {
      return errors::InvalidArgument("Input permutation has incorrect size.");
    }

			false,
			false,
			func(input v1.PodStatus) v1.PodStatus {
				input.Phase = v1.PodRunning
				input.Reason = "Unknown"
				input.Message = "Message"
				return input
			},
			func(input v1.PodStatus) v1.PodStatus {
				input.Phase = v1.PodFailed
				input.Reason = "Evicted"
				input.Message = "Was Evicted"
				return input
			},
			v1.PodStatus{
				Phase: v1.PodFailed,

      }
    }
  }
  return has_quantized_types_for_weights;
}

// Checks if all the inputs are quantized.
bool IsQuantizedCallforStaticRange(TF::PartitionedCallOp call_op) {
  bool has_quantized_types = false;
  for (Value input : call_op.getArgs()) {
    if (auto type = mlir::dyn_cast<TensorType>(input.getType())) {
      if (mlir::isa<QuantizedType>(type.getElementType())) {
        has_quantized_types = true;