Corefunctions

Corefunctionality for data preparation of sequential data for pytorch, fastai models

5.3 Spectrogram-Datablock

TensorSpectrogramOutput

 TensorSpectrogramOutput (x, **kwargs)

A Tensor which support subclass pickling, and maintains metadata when casting or after methods

source

TensorSpectrogramInput

 TensorSpectrogramInput (x, **kwargs)

A Tensor which support subclass pickling, and maintains metadata when casting or after methods

source

TensorSpectrogram

 TensorSpectrogram (x, **kwargs)

A Tensor which support subclass pickling, and maintains metadata when casting or after methods

from nbdev.config import get_config

project_root = get_config().config_file.parent
f_path = project_root / 'test_data/WienerHammerstein'
hdf_files = get_files(f_path,extensions='.hdf5',recurse=True)

seq_spec = TensorSpectrogramInput(HDF2Sequence(['u'],to_cls=TensorSpectrogramInput)._hdf_extract_sequence(hdf_files[0],r_slc=2000))
seq_spec.shape

torch.Size([2000, 1])

source

Sequence2Spectrogram

 Sequence2Spectrogram (scaling='log', n_fft:int=400,
                       win_length:Optional[int]=None,
                       hop_length:Optional[int]=None, pad:int=0,
                       window_fn:Callable[...,torch.Tensor]=<built-in
                       method hann_window of type object at
                       0x7fa5c7f36c20>, power:Optional[float]=2.0,
                       normalized:bool=False, wkwargs:Optional[dict]=None,
                       **kwargs)

calculates the FFT of a sequence

source

SpectrogramBlock

 SpectrogramBlock (seq_extract, padding=False, n_fft=100, hop_length=None,
                   normalized=False)

A basic wrapper that links defaults transforms for the data block API

dls_spec = DataBlock(blocks=(SpectrogramBlock.from_hdf(['u','y'],n_fft=100,hop_length=10,normalized=True),
                             SequenceBlock.from_hdf(['y'],TensorSequencesOutput)),
                get_items= CreateDict([DfHDFCreateWindows(win_sz=2000+1,stp_sz=10,clm='u')]),
                splitter=ApplyToDict(ParentSplitter())).dataloaders(hdf_files)

dls_spec.one_batch()[0].shape

torch.Size([64, 1, 51, 201])

dls_spec.show_batch(max_n=1)