[CUDA-enabled GPU cards](https://www.tensorflow.org/install/gpu) *(Ubuntu and
Windows)*:

```
$ pip install tensorflow
```

Other devices (DirectX and MacOS-metal) are supported using
[Device Plugins](https://www.tensorflow.org/install/gpu_plugins#available_devices).

A smaller CPU-only package is also available:

```
$ pip install tensorflow-cpu
```

* Fix issue where on re-registration of device plugin, `allocatable` was not getting updated. This issue makes devices invisible to the Kubelet if device plugin restarts. Only work-around, if this fix is not there, is to restart the kubelet and then start device plugin. ([#63118](https://github.com/kubernetes/kubernetes/pull/63118), [@vikaschoudhary16](https://github.com/vikaschoudhary16))

    ask_cuda_compute_capabilities = (
        'Please specify a list of comma-separated CUDA compute capabilities '
        'you want to build with.\nYou can find the compute capability of your '
        'device at: https://developer.nvidia.com/cuda-gpus. Each capability '
        'can be specified as "x.y" or "compute_xy" to include both virtual and'
        ' binary GPU code, or as "sm_xy" to only include the binary '

# Handling Errors { #handling-errors }

There are many situations in which you need to notify an error to a client that is using your API.

This client could be a browser with a frontend, a code from someone else, an IoT device, etc.

You could need to tell the client that:

* The client doesn't have enough privileges for that operation.
* The client doesn't have access to that resource.
* The item the client was trying to access doesn't exist.
* etc.

## Add Labels to PRs

The same GitHub Action <a href="https://github.com/tiangolo/latest-changes" class="external-link" target="_blank">latest-changes</a> uses one label in the PR to decide the section in the release notes to put this PR in.

Make sure you use a supported label from the <a href="https://github.com/tiangolo/latest-changes#using-labels" class="external-link" target="_blank">latest-changes list of labels</a>:

 * UTF-8.
 *
 * <p><b>Note:</b> This escaper produces <a
 * href="https://url.spec.whatwg.org/#percent-encode">uppercase</a> hexadecimal sequences.
 *
 * @author David Beaumont
 * @since 15.0
 */
@GwtCompatible
public final class PercentEscaper extends UnicodeEscaper {

  // In some escapers spaces are escaped to '+'
  private static final char[] plusSign = {'+'};

 * to create multiple escaper instances that have the same character replacement mapping consider
 * using {@link ArrayBasedEscaperMap}.
 *
 * @author David Beaumont
 * @since 15.0
 */
@GwtCompatible
@SuppressWarnings("EscapedEntity") // We do mean for the user to see "&" etc.
public abstract class ArrayBasedUnicodeEscaper extends UnicodeEscaper {

* Backporting TPR fix to 1.4 ([#42380](https://github.com/kubernetes/kubernetes/pull/42380), [@foxish](https://github.com/foxish))
* Fix AWS device allocator to only use valid device names ([#41455](https://github.com/kubernetes/kubernetes/pull/41455), [@gnufied](https://github.com/gnufied))

package main

import (
	"fmt"
	"go/build"
	"log"
	"os"
	"os/exec"
	"path/filepath"
	"runtime"
	"strings"
	"syscall"
)

const debug = false

var tmpdir string

var (
	devID    string
	appID    string
	teamID   string
	bundleID string
	deviceID string
)

// lock is a file lock to serialize iOS runs. It is global to avoid the

to compute (y * q) / 2ᵈ, rounded to nearest integer, with 1/2 // rounding up (see FIPS 203, Section 2.3). dividend := uint32(y) * q quotient := dividend >> d // (y * q) / 2ᵈ // The d'th least-significant bit of the dividend (the most significant bit // of the remainder) is 1 for the top half of the values that divide to the // same quotient, which are the ones that round up. quotient += dividend >> (d - 1) & 1 // quotient is at most (2¹¹-1) * q / 2¹¹ + 1 = 3328, so it didn't overflow. return fieldElement(quotient)...