Dataset<'S>
Namespace: Datasets
A dataset of a record type 'S containing ArrayNDT<_> data variables. The first dimension of each record field is the sample. All record fields must contain the same number of samples. The default constructor expects a list of ArrayNDTs corresponding to the fields in 'S. To construct a Dataset<_> from a sequence of samples use the Dataset<_>.FromSamples method.
Constructors
| Constructor | Description |
new(fieldStorages)
Signature: (fieldStorages:ITensor list) -> Dataset<'S>
|
Creates a non-sequence dataset using the specified field storages. CompiledName: |
new(fieldStorages, isSeq)
Signature: (fieldStorages:ITensor list * isSeq:bool) -> Dataset<'S>
|
CompiledName: |
Instance members
| Instance member | Description |
All
Signature: 'S
|
Returns a record of type 'S containing all samples. CompiledName: |
Batches(batchSize)
Signature: batchSize:int64 -> seq<'S>
|
Returns a sequence of batches with the given size of this dataset. If the number of samples in this dataset is not a multiple of the batch size, the last batch will be smaller. |
FieldStorages
Signature: ITensor list
|
list of arrays corresponding to the data of each field in the sample record CompiledName: |
GetSlice(...)
Signature: (startSmpl:int64 option * stopSmpl:int64 option * startStep:int64 option * stopStep:int64 option) -> 'S
|
For a sequence dataset, returns a record of type 'S containing a slice of samples and slots. |
GetSlice(start, stop)
Signature: (start:int64 option * stop:int64 option) -> 'S
|
Returns a record of type 'S containing a slice of samples. |
IsSeq
Signature: bool
|
true if this is a sequence dataset CompiledName: |
[smpl]
Signature: smpl:int64 -> 'S
|
Returns a record of type 'S containing the sample with the given index. CompiledName: |
[smpl, step]
Signature: smpl:int64 * step:int64 -> 'S
|
For a sequence dataset, returns a record of type 'S containing the sample and slot with the given indices. CompiledName: |
Location
Signature: ITensorDevice
|
storage location CompiledName: |
NSamples
Signature: int64
|
number of samples CompiledName: |
NSteps
Signature: int64
|
number of steps for sequence dataset CompiledName: |
PaddedBatches(batchSize)
Signature: batchSize:int64 -> unit -> seq<'S>
|
Generates a function that returns a sequence of batches with the given size of this dataset. If the number of samples in this dataset is not a multiple of the batch size, the last batch will still have the specified size but is padded with zeros. |
Part(start, stop)
Signature: (start:int64 * stop:int64) -> Dataset<'S>
|
Returns a new dataset containing the samples from start to stop. |
Pretty
Signature: string
|
pretty string CompiledName: |
SampleType
Signature: Type
|
data type of samples CompiledName: |
Save(filename)
Signature: filename:string -> unit
|
Saves this dataset into the specified HDF5 file. The file is overwritten. |
Save(hdf, hdfPrefixPath)
Signature: (hdf:HDF5 * hdfPrefixPath:string option) -> unit
|
Saves this dataset into the specified HDF5 file.
|
SlotBatches(batchSize slotSize)
Signature: batchSize:int64 -> slotSize:int64 -> seq<'S>
|
Returns a sequence of batches of time slots with size |
TmplBatch(batchSize)
Signature: batchSize:int64 -> 'S
|
template batch |
Static members
| Static member | Description |
Load(filename)
Signature: filename:string -> Dataset<'S>
|
Loads a dataset from the specified HDF5 file. |
Load(hdf, hdfPrefixPath)
Signature: (hdf:HDF5 * hdfPrefixPath:string option) -> Dataset<'S>
|
Loads a dataset from the specified HDF5 file.
|
New(samples, isSeq)
Signature: (samples:seq<'S> * isSeq:bool) -> Dataset<'S>
|
Constructs a dataset from samples. |
