int64_t& inferred_dim = (*inferred_item_shape)[i];
      int64_t dim = item_shape[i];
      if (ShapedType::isDynamic(inferred_dim)) inferred_dim = dim;
    }
  }

  // If all indices are constants, then verify that they cover all indices in
  // the range [0, max_index] and the output type is legal.
  if (all_indices_const) {
    for (int32_t i = 0; i <= max_index; i++) {
      if (!index_values.count(i))

	// 0-indexed names, or to orchestrate progressive movement of replicas from
	// one StatefulSet to another.
	// If set, replica indices will be in the range:
	//   [.spec.ordinals.start, .spec.ordinals.start + .spec.replicas).
	// If unset, defaults to 0. Replica indices will be in the range:
	//   [0, .spec.replicas).
	// +optional
	Start int32 `json:"start" protobuf:"varint,1,opt,name=start"`
}

	s.stksize = 0
	s.stkptrsize = 0
	s.stkalign = int64(types.RegSize)
	fn := s.curfn

	// Mark the PAUTO's unused.
	for _, ln := range fn.Dcl {
		if ln.OpenDeferSlot() {
			// Open-coded defer slots have indices that were assigned
			// upfront during SSA construction, but the defer statement can
			// later get removed during deadcode elimination (#61895). To
			// keep their relative offsets correct, treat them all as used.

// CHECK:           %[[VALUES:.*]], %[[INDICES:.*]] = "tfl.topk_v2"(%[[ARG]], %[[K]]) : (tensor<3x10xf32>, tensor<i32>) -> (tensor<3x5xf32>, tensor<3x5xi32>)
// CHECK:           return %[[INDICES]] : tensor<3x5xi32>
// CHECK:         }
func.func @optimizeTopKOnlyIndices(%arg: tensor<3x10xf32>) -> tensor<3x5xi32>{
  %K = arith.constant dense<10> : tensor<i32>

  // inferred from the tensor content and op property. A "-1" value indicates
  // the operand doesn't support per-channel quantization.
  llvm::DenseMap<int, int> coeff_op_quant_dim;

  // Indices of quantizable operands. Biases are not included in this field,
  // the indices of biases can be found in the `biases_params`.
  absl::flat_hash_set<int> quantizable_operands;
};

        if self.same_scale_op == 'concatenate':
          ones = array_ops.ones_like(out)
          out = array_ops.concat([out, ones], 0)
        elif self.same_scale_op == 'gather':
          out = array_ops.gather(out, indices=[0], axis=0)
        elif self.same_scale_op == 'max_pool':
          out = nn_ops.max_pool(out, ksize=3, strides=1, padding='SAME')
        elif self.same_scale_op == 'pad':
          paddings = array_ops.ones(

		driverResources = active.resources
	}
	c.mutex.RUnlock()

	// Resources that are not yet stored in any slice need to be published.
	// Here we track the indices of any resources that are already stored.
	storedResourceIndices := sets.New[int]()

	// Slices that don't match any driver resource can either be updated (if there

labels.maintenance_title_configuration=Maintenance
labels.number_of_shards_for_doc=The number of shards
labels.auto_expand_replicas_for_doc=Auto expand replicas
labels.clear_crawler_index=Crawler Indices
labels.clear_crawler_index_button=Clear Crawler Indices
labels.diagnostic_logs=Diagnostic
labels.download_diagnostic_logs_button=Download Logs
labels.reload_doc_index=Reload Doc Index
labels.reload_doc_index_button=Reload

labels.maintenance_title_configuration=Maintenance
labels.number_of_shards_for_doc=The number of shards
labels.auto_expand_replicas_for_doc=Auto expand replicas
labels.clear_crawler_index=Crawler Indices
labels.clear_crawler_index_button=Clear Crawler Indices
labels.diagnostic_logs=Diagnostic
labels.download_diagnostic_logs_button=Download Logs
labels.reload_doc_index=Reload Doc Index
labels.reload_doc_index_button=Reload

   * equivalent to RegularImmutableSet, save that instead of having a hash table containing the
   * elements directly and null for empty positions, we store indices of the keys in the hash table,
   * and ABSENT for empty positions.  We then look up the keys in alternatingKeysAndValues.
   *
   * (The index actually stored is the index of the key in alternatingKeysAndValues, which is