}
}

// Runs model as is if output is a scalar,
// else sums the output tensor before returning.
Status RunAndMaybeSum(AbstractContext* ctx, Model forward,
                      absl::Span<AbstractTensorHandle* const> inputs,
                      absl::Span<AbstractTensorHandle*> outputs,
                      bool use_function) {
  AbstractTensorHandle* model_outputs[1];

  // Run the model.
  TF_RETURN_IF_ERROR(

  Status ComputeGradient(
      const VSpace<Gradient, BackwardFunction, TapeTensor>& vspace,
      const absl::Span<const int64_t> target_tensor_ids,
      const absl::Span<const int64_t> source_tensor_ids,
      const std::unordered_map<int64, TapeTensor>& sources_that_are_targets,
      gtl::ArraySlice<Gradient*> output_gradients, absl::Span<Gradient*> result,
      bool build_default_zeros_grads = true);

### 实时文档

在本地开发时，可以使用该脚本构建站点并检查所做的任何更改，并实时重载：

<div class="termy">

```console
$ python ./scripts/docs.py live

<span style="color: green;">[INFO]</span> Serving on http://127.0.0.1:8008
<span style="color: green;">[INFO]</span> Start watching changes
<span style="color: green;">[INFO]</span> Start detecting changes
```

</div>

文档服务将运行在 `http://127.0.0.1:8008`。

这样，你可以编辑文档 / 源文件并实时查看更改。

!!! tip

    <a class="badge" href="https://dpeuniversity.gradle.com/app/courses/ec69d0b8-9171-4969-ac3e-82dea16f87b0/" target="_blank">
        <span class="badge-type button--blue">LEARN</span>
        <span class="badge-text">Incremental Builds and Build Caching with Gradle&nbsp;&nbsp;&nbsp;&gt;</span>
    </a>
</div>
++++

Gradle uses two main features to reduce build time: *incremental builds* and *build caching*.

    <a class="badge" href="https://dpeuniversity.gradle.com/app/courses/ec69d0b8-9171-4969-ac3e-82dea16f87b0/" target="_blank">
        <span class="badge-type button--blue">LEARN</span>
        <span class="badge-text">Incremental Builds and Build Caching with Gradle&nbsp;&nbsp;&nbsp;&gt;</span>
    </a>
</div>
++++

Learn the basics of Gradle's incremental builds.

****
**In this section you will:**

			continue
		}
		obj, span, objIndex := findObject(ptr, 0, 0)
		if obj == 0 {
			continue
		}
		// TODO: Consider making two passes where the first
		// just prefetches the mark bits.
		mbits := span.markBitsForIndex(objIndex)
		if mbits.isMarked() {
			continue
		}
		mbits.setMarked()

		// Mark span.
		arena, pageIdx, pageMask := pageIndexOf(span.base())

    params->emplace_back(handle);
  }
  return absl::OkStatus();
}

// Runs `model` maybe wrapped in a function.
Status RunModel(Model model, AbstractContext* ctx,
                absl::Span<AbstractTensorHandle* const> inputs,
                absl::Span<AbstractTensorHandle*> outputs, bool use_function) {
  if (use_function) {
    const char* fn_name = "test_fn";
    core::RefCountPtr<AbstractFunction> scoped_func;

#include "tensorflow/core/lib/core/status_test_util.h"
#include "tensorflow/core/platform/test.h"

namespace tensorflow {
namespace {

Status PickDeviceHelper(bool allow_mixing_unknown_and_cpu,
                        absl::Span<const absl::string_view> device_names,
                        string* result) {
  jit::DeviceInfoCache cache;
  jit::DeviceSet device_set;
  for (absl::string_view name : device_names) {

#include "tensorflow/compiler/mlir/quantization/tensorflow/calibrator/calibration_statistics_collector_min_max.h"

#include <algorithm>
#include <cstdint>
#include <limits>
#include <optional>

#include "absl/types/span.h"
#include "tensorflow/compiler/mlir/quantization/tensorflow/calibrator/calibration_statistics.pb.h"

namespace tensorflow {
namespace calibrator {

void CalibrationStatisticsCollectorMinMax::ClearData() {

  @Override
  public ImmutableRangeSet<C> subRangeSet(Range<C> range) {
    if (!isEmpty()) {
      Range<C> span = span();
      if (range.encloses(span)) {
        return this;
      } else if (range.isConnected(span)) {
        return new ImmutableRangeSet<>(intersectRanges(range));
      }
    }
    return of();
  }

  /**