// once unless the tape is persistent) and produces the gradient of the target
  // tensors with respect to the source tensors. The output gradients are used
  // if not empty and not null. The result is populated with one tensor per
  // target element.
  absl::Status ComputeGradient(
      AbstractContext* ctx, absl::Span<AbstractTensorHandle* const> targets,
      absl::Span<AbstractTensorHandle* const> sources,

## Configure Prometheus to use AlertManager

Add below section to your `prometheus.yml`
```yaml
alerting:
  alertmanagers:
  - static_configs:
    - targets: ['localhost:9093']
rule_files:
  - rules.yml
```
Here `rules.yml` is the file which should contain the alerting rules defined.

## Add rules for your deployment

	Flags      BatchJobReplicateFlags  `yaml:"flags" json:"flags"`
	Target     BatchJobReplicateTarget `yaml:"target" json:"target"`
	Source     BatchJobReplicateSource `yaml:"source" json:"source"`

	clnt *miniogo.Core `msg:"-"`
}

// RemoteToLocal returns true if source is remote and target is local
func (r BatchJobReplicateV1) RemoteToLocal() bool {
	return !r.Source.Creds.Empty()

# START MACOS CROSS-COMPILE CONFIGS
build:cross_compile_macos_x86 --config=cross_compile_base
build:cross_compile_macos_x86 --config=nonccl
# Target Catalina (10.15) as the minimum supported OS
build:cross_compile_macos_x86 --action_env  MACOSX_DEPLOYMENT_TARGET=10.15

# Set the target CPU to Darwin x86
build:cross_compile_macos_x86 --platforms=//tensorflow/tools/toolchains/cross_compile/config:darwin_x86_64

  void SetOutputs(std::initializer_list<TF_Operation*> outputs);
  void SetOutputs(const std::vector<TF_Output>& outputs);
  void SetTargets(std::initializer_list<TF_Operation*> targets);

  void Run(TF_Status* s);

  void CloseAndDelete(TF_Status* s);

  TF_Tensor* output_tensor(int i) { return output_values_[i]; }

  TF_Session* mutable_session() { return session_; }

func (d Destination) LegacyArn() bool {
	return strings.HasPrefix(d.ARN, DestinationARNPrefix)
}

// TargetArn returns true if arn format has prefix  "arn:minio:replication:::"
// used for multi-destination targets
func (d Destination) TargetArn() bool {
	return strings.HasPrefix(d.ARN, DestinationARNMinIOPrefix)
}

// MarshalXML - encodes to XML data.
func (d Destination) MarshalXML(e *xml.Encoder, start xml.StartElement) error {

    git config --system --add safe.directory '*'
    git config --system protocol.file.allow always

    # Note that you could generate a list of all the affected targets with e.g.:
    # bazel query $(paste -sd "+" $BATS_FILE_TMPDIR/changed_files) --keep_going
    # Only shows Added, Changed, Modified, Renamed, and Type-changed files

)

# To avoid duplication, check that the C++ or python library does not depend on
# the stream executor cuda plugins. Targets that want to use cuda APIs should
# instead depend on the dummy plugins in @local_tsl//tsl/platform/default/build_config
# and use header only targets.
# TODO(ddunleavy): This seems completely broken. :tensorflow_cc depends on
# cuda_platform from tf_additional_binary_deps and this doesn't break.

MINIO_API_OBJECT_MAX_VERSIONS             (number)    set max allowed number of versions per object (default: '9223372036854775807')
```

#### Notifications

Notification targets supported by MinIO are in the following list. To configure individual targets please refer to more detailed documentation [here](https://min.io/docs/minio/linux/administration/monitoring.html#bucket-notifications).

```

Replication from a source bucket to multiple destination buckets is supported. For each of the targets, repeat the steps to configure a remote target ARN and add replication rules to the source bucket's replication config.