// -1 and `dims` can be null. If a dimension is unknown, the
// corresponding entry in the `dims` array must be -1.
//
// This does not overwrite the existing shape associated with `output`,
// but merges the input shape with the existing shape.  For example,
// setting a shape of [-1, 2] with an existing shape [2, -1] would set
// a final shape of [2, 2] based on shape merging semantics.
//

        # GPU:
        *manylinux*)     LARGEST_OK_SIZE=580 ;;
        # Unknown:
        *)
            echo "The wheel's name is in an unknown format."
            exit 1
            ;;
    esac
    # >&3 forces output in bats even if the test passes. See
    # https://bats-core.readthedocs.io/en/stable/writing-tests.html#printing-to-the-terminal
    echo "# Size of $TF_WHEEL is $WHEEL_MEGABYTES / $LARGEST_OK_SIZE megabytes." >&3

> directory are stable and appropriate documentation around its usage is in
> place.

Maintainer: TensorFlow DevInfra

Issue Reporting: File an issue against this repo and tag
[@devinfra](https://github.com/orgs/tensorflow/teams/devinfra)

********************************************************************************

A directory for build and CI related scripts and jobs managed by the TensorFlow

  if (op_reg_data->shape_inference_fn == nullptr) {
    status->status =
        InvalidArgument("No shape inference function exists for op '",
                        node_def.op(), "', did you forget to define it?");
    return;
  }

  status->status = c.Run(op_reg_data->shape_inference_fn);
  if (!status->status.ok()) return;

  // Set output_shapes.
  TF_ShapeAndTypeList* output_shapes_result =

# Official CI Directory

Maintainer: TensorFlow and TensorFlow DevInfra

Issue Reporting: File an issue against this repo and tag
[@devinfra](https://github.com/orgs/tensorflow/teams/devinfra)

********************************************************************************

## TensorFlow's Official CI and Build/Test Scripts

TensorFlow's official CI jobs run the scripts in this folder. Our internal CI

  EXPECT_TRUE(executed);
  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());
  TFE_DeleteTensorHandle(retval);

  // Explicit placement still forces the op onto the requested device
  matmul.reset(MatMulOp(context.get(), hcustom0.get(), hcpu.get()));
  TFE_OpSetDevice(matmul.get(), "/job:localhost/replica:0/task:0/device:CPU:0",
                  status.get());

          (std::min)(n - copy_size, gcs_file->buffer.size());
      memcpy(buffer + copy_size, gcs_file->buffer.data(), remaining_copy);
      copy_size += remaining_copy;
    }
    if (copy_size < n) {
      // Forget the end-of-file flag to allow for clients that poll on the
      // same file.
      gcs_file->buffer_end_is_past_eof = false;
      TF_SetStatus(status, TF_OUT_OF_RANGE, "Read less bytes than requested");

    `float16` for acceleration on Volta and Turing Tensor Cores. This feature
    can be enabled by wrapping an optimizer class with
    `tf.train.experimental.enable_mixed_precision_graph_rewrite()`.
*   Add environment variable `TF_CUDNN_DETERMINISTIC`. Setting to "true" or "1"
    forces the selection of deterministic cuDNN convolution and max-pooling