url: https://github.com/BoostryJP
  - login: acsone
    avatarUrl: https://avatars.githubusercontent.com/u/7601056?v=4
    url: https://github.com/acsone
- - login: scalar
    avatarUrl: https://avatars.githubusercontent.com/u/301879?v=4
    url: https://github.com/scalar
  - login: Trivie
    avatarUrl: https://avatars.githubusercontent.com/u/8161763?v=4
    url: https://github.com/Trivie
- - login: takashi-yoneya

namespace tensorflow {

// PyCapsule name for DLPack Tensor
const char* const kDlTensorCapsuleName = "dltensor";

// Returns the DLDevice* for the given eager tensor handle.
//
// The caller takes ownership of the returned pointer and is responsible for
// deleting it.
TF_CAPI_EXPORT extern void* TFE_GetDLDevice(TFE_TensorHandle* h,
                                            TF_Status* status);

     * This method does interpret tool arguments.
     *
     * @param parserRequest the request containing all necessary information for parsing
     * @return the parsed invoker request. Caller must start by checking {@link InvokerRequest#parsingFailed()} as
     * if there are parser errors, this request may not be fully processed and should immediately be failed.
     */
    @Nonnull

  }

  final Set<Throwable> getOrInitSeenExceptions() {
    /*
     * The initialization of seenExceptionsField has to be more complicated than we'd like. The
     * simple approach would be for each caller CAS it from null to a Set populated with its
     * exception. But there's another race: If the first thread fails with an exception and a second
     * thread immediately fails with the same exception:
     *

    /**
     * Share capability indicating continuous availability support.
     */
    public static final int SMB2_SHARE_CAP_CONTINUOUS_AVAILABILITY = 0x10;

    /**
     * Share capability indicating scale-out support.
     */
    public static final int SMB2_SHARE_CAP_SCALEOUT = 0x20;

    /**
     * Share capability indicating cluster support.
     */
    public static final int SMB2_SHARE_CAP_CLUSTER = 0x40;

 *
 * Other exception types cancel the current call:
 *
 *  * For synchronous calls made with [Call.execute], the exception is propagated to the caller.
 *
 *  * For asynchronous calls made with [Call.enqueue], an [IOException] is propagated to the caller
 *    indicating that the call was canceled. The interceptor's exception is delivered to the current

`Policy` optionally took in a `tf.compat.v1.mixed_precision.LossScale` in the constructor, which defaulted to a dynamic loss scale for the `"mixed_float16"` policy and no loss scale for other policies. In `Model.compile`, if the model's policy had a loss scale, the optimizer would be wrapped with a `LossScaleOptimizer`. With the non-experimental `Policy`, there is no loss scale associated with the `Policy`, and `Model.compile` wraps the optimizer with a `LossScaleOptimizer` if and only if the policy...

 */
@Deprecated(since = "4.0.0")
public interface Source {

    /**
     * Gets a byte stream to the source contents. Closing the returned stream is the responsibility of the caller.
     *
     * @return A byte stream to the source contents, never {@code null}.
     * @throws IOException in case of IO issue
     */
    InputStream getInputStream() throws IOException;

    /**

 */
@Deprecated(since = "4.0.0")
public interface SuperPomProvider {

    /**
     * Gets the super POM for the specified model version. The returned model is supposed to be read-only, i.e. if the
     * caller intends to make updates to the model the return value must be cloned before updating to ensure the
     * modifications don't affect future retrievals of the super POM.
     *

streaming compression due to its stability and performance.

This algorithm is specifically optimized for machine generated content.
Write throughput is typically at least 500MB/s per CPU core,
and scales with the number of available CPU cores.
Decompression speed is typically at least 1GB/s.

This means that in cases where raw IO is below these numbers