draugr.torch_utilities.writers.tensorboard.tensorboard_pytorch_writer.TensorBoardPytorchWriter

class draugr.torch_utilities.writers.tensorboard.tensorboard_pytorch_writer.TensorBoardPytorchWriter(path: Union[str, Path] = PosixPath('/mnt/win/Users/Christian/ProjectsWin/Github/Pything/draugr/docs/Logs'), summary_writer_kws=None, *, interval: Optional[int] = 1, filters: Iterable = None, verbose: bool = False)[source]

Bases: Writer, ImageWriterMixin, GraphWriterMixin, HistogramWriterMixin, BarWriterMixin, LineWriterMixin, SpectrogramWriterMixin, FigureWriterMixin, InstantiationWriterMixin, PrecisionRecallCurveWriterMixin, EmbedWriterMixin, ModuleParameterWriterMixin, VideoWriterMixin, MeshWriterMixin

Provides a pytorch-tensorboard-implementation writer interface

__init__(path: Union[str, Path] = PosixPath('/mnt/win/Users/Christian/ProjectsWin/Github/Pything/draugr/docs/Logs'), summary_writer_kws=None, *, interval: Optional[int] = 1, filters: Iterable = None, verbose: bool = False)[source]
Parameters
  • path

  • summary_writer_kws

  • kwargs

Methods

__init__([path, summary_writer_kws, ...])

param path

bar(tag, values, step[, value_error, ...])

param x_label

blip(tag[, step_i])

param tag

close()

description

embed(tag, response[, metadata, label_img, ...])

BORKED!

figure(tag, figure, step, *[, global_step, ...])

param tag

filter(tag)

returns a boolean value, true if to be included, False if to be excluded

graph(model, input_to_model[, verbose, ...])

param use_strict_trace

histogram(tag, values, step[, bins, ...])

param tag

image(tag, data, step, *[, data_formats, ...])

param self

instance(instance, metrics, **kwargs)

TODO: Not finished!

line(tag, values, step[, x_labels, y_label, ...])

param x_label

mesh(tag, data[, step])

Data being vertices here.

open()

description

parameters(model, step[, tag])

param model

precision_recall_curve(tag, predictions, truths)

param tag

scalar(tag, value[, step_i])

param tag

spectrogram(tag, values, sample_rate, step)

param sample_rate

video(tag, data[, step, frame_rate, input_dims])

Shape:

Attributes

writer

return: :rtype:

class MultiChannelMethodEnum(value)

Bases: Enum

An enumeration.

bar(tag: str, values: list, step: int, value_error: float = None, x_labels: Sequence = None, y_label: str = 'Probabilities', x_label: str = 'Distribution', *, global_step=None, close=True, walltime=None) None[source]
Parameters
  • x_label

  • tag

  • values

  • step

  • value_error

  • x_labels

  • y_label

  • kwargs

blip(tag: str, step_i: Optional[int] = None) None
Parameters
  • tag

  • step_i

close() Any

description

embed(tag: str, response: Sequence, metadata: Optional[Any] = None, label_img: Optional[Any] = None, step: Optional[int] = None, *, global_step=None, metadata_header=None) None[source]

BORKED!

Parameters
  • tag

  • response

  • metadata

  • label_img

  • step

  • kwargs

Returns

figure(tag: str, figure: Figure, step: int, *, global_step=None, close=True, walltime=None) None[source]
Parameters
  • tag

  • figure

  • step

  • kwargs

filter(tag: str) bool

returns a boolean value, true if to be included, False if to be excluded

tag is in filter if not None and within interval for inclusion

Parameters

tag

Returns

Return type

graph(model: Module, input_to_model: Tensor, verbose: bool = False, use_strict_trace: bool = True) None[source]
Parameters
  • use_strict_trace

  • verbose

  • model

  • input_to_model

histogram(tag: str, values: list, step: int, bins: Any = 'auto', *, global_step=None, walltime=None, max_bins=None) None[source]
Parameters
  • tag

  • values

  • step

  • bins

  • kwargs

image(tag: str, data: Union[ndarray, Tensor, Image], step, *, data_formats: str = 'NCHW', multi_channel_method: MultiChannelMethodEnum = MultiChannelMethodEnum.seperate, seperate_channel_postfix: str = 'channel', seperate_image_postfix: str = 'image', global_step=None, walltime=None, dataformats='CHW') None[source]
Parameters
  • self

  • multi_channel_method

  • seperate_channel_postfix

  • seperate_image_postfix

  • tag

  • data

  • step

  • data_formats

  • kwargs

instance(instance: dict, metrics: dict, **kwargs) None[source]

TODO: Not finished!

Parameters
  • instance

  • metrics

Returns

line(tag: str, values: list, step: int, x_labels: Sequence = None, y_label: str = 'Magnitude', x_label: str = 'Sequence', plot_kws=None, *, global_step=None, close=True, walltime=None) None[source]
Parameters
  • x_label

  • plot_kws

  • tag

  • values

  • step

  • x_labels

  • y_label

  • kwargs

mesh(tag: str, data: Union[ndarray, Tensor, Image], step=None, **kwargs) None[source]

Data being vertices here.

Shape: (B,N,3) data: (B,N,3). (batch, number_of_vertices, channels) colors: (B,N,3). The values should lie in [0, 255] for type uint8 or [0, 1] for type float. faces: (B,N,3). The values should lie in [0, number_of_vertices] for type uint8.

Parameters
  • tag

  • data

  • step

  • kwargs

Returns

open() Any

description

parameters(model: Module, step: int, tag: str = '', **kwargs) None[source]
Parameters
  • model

  • step

  • tag

  • kwargs

precision_recall_curve(tag: str, predictions: Iterable, truths: Iterable, step: int = None, *, global_step=None, num_thresholds=127, weights=None, walltime=None) None[source]
Parameters
  • tag

  • predictions

  • truths

  • step

  • kwargs

scalar(tag: str, value: Union[int, float], step_i: Optional[int] = None) None
Parameters
  • tag

  • value

  • step_i

spectrogram(tag: str, values: list, sample_rate: int, step: int, n_fft: int = 512, step_size=128, x_labels: Sequence = None, y_label: str = 'Frequency [Hz]', x_label: str = 'Time [sec]', plot_kws=None, *, global_step=None, close=True, walltime=None) None[source]
Parameters
  • sample_rate

  • n_fft

  • step_size

  • tag

  • values

  • step

  • x_labels

  • y_label

  • x_label

  • plot_kws

  • kwargs

video(tag: str, data: Union[ndarray, Tensor, Image], step=None, frame_rate=30, input_dims=VideoInputDimsEnum.ntchw, **kwargs) None[source]

Shape:

fastest expects vid_tensor: (N,T,C,H,W) .

The values should lie in [0, 255] for type uint8 or [0, 1] for type float.

property writer: SummaryWriter

return: :rtype: