all_valid = False
        else:
          ver = int(sm_compute_match.group(2))
          if ver < 30:
            print(
                'ERROR: TensorFlow only supports small CUDA compute'
                ' capabilities of sm_30 and higher. Please re-specify the list'
                ' of compute capabilities excluding version %s.' % ver)
            all_valid = False
          if ver < 35:

// compacted until the limit is reached.
//
// This ensures that any branch will never contain an unreasonable amount of other branches,
// and also that small branches with few objects don't take up unreasonable amounts of space.
// This keeps the cache size at a reasonable size for all buckets.
//
// Whenever a branch is scanned, it is assumed that it will be un-compacted

   1. Definitions.

      "License" shall mean the terms and conditions for use, reproduction,
      and distribution as defined by Sections 1 through 9 of this document.

      "Licensor" shall mean the copyright owner or entity authorized by
      the copyright owner that is granting the License.

      "Legal Entity" shall mean the union of the acting entity and all

  double stableIntervalMicros;

  /**
   * The time when the next request (no matter its size) will be granted. After granting a request,
   * this is pushed further in the future. Large requests push this further than small requests.
   */
  private long nextFreeTicketMicros = 0L; // could be either in the past or future

  private SmoothRateLimiter(SleepingStopwatch stopwatch) {
    super(stopwatch);
  }

  @Override

* The title is correct following the instructions above.
* It has the labels `lang-all` and `lang-{lang code}`.
* The PR changes only one Markdown file adding a translation.
    * Or in some cases, at most two files, if they are small, for the same language, and people reviewed them.
    * If it's the first translation for that language, it will have additional `mkdocs.yml` files, for those cases follow the instructions below.

//     backed by this underlying array.
//   - Map: An empty map is allocated with enough space to hold the
//     specified number of elements. The size may be omitted, in which case
//     a small starting size is allocated.
//   - Channel: The channel's buffer is initialized with the specified
//     buffer capacity. If zero, or the size is omitted, the channel is
//     unbuffered.

    // as their length. These files are very small, so it's wasteful to allocate an 8KB buffer.
    int initialBufferSize = min(BUFFER_SIZE, max(128, Integer.highestOneBit(totalLen) * 2));
    // Starting with an 8k buffer, double the size of each successive buffer. Smaller buffers
    // quadruple in size until they reach 8k, to minimize the number of small reads for longer

     */
    @Nonnull
    default Path targetPath(@Nonnull Project project) {
        Optional<Path> targetPath = targetPath();
        // The test for `isAbsolute()` is a small optimization for avoiding the call to `getOutputDirectory(…)`.
        return targetPath.filter(Path::isAbsolute).orElseGet(() -> {
            Path base = project.getOutputDirectory(scope());

In **SQLModel**, any model class that has `table=True` is a **table model**.

And any model class that doesn't have `table=True` is a **data model**, these ones are actually just Pydantic models (with a couple of small extra features). 🤓

With SQLModel, we can use **inheritance** to **avoid duplicating** all the fields in all the cases.

#### `HeroBase` - the base class { #herobase-the-base-class }

И мы, как разработчики, продолжаем улучшать код — находим баги и добавляем новые возможности (иногда добавляя новые баги 😅).

### Небольшие ошибки обрабатываются автоматически { #small-errors-automatically-handled }

Создавая веб‑API с FastAPI, если в нашем коде возникает ошибка, FastAPI обычно «локализует» её в пределах одного запроса, который эту ошибку вызвал. 🛡