1.5. RectiPy.rectipy.observer

Module that contains rectipy.observer.Observer, the class that is used in RectiPy to store states of the rectipy.Network.rnn_layer.

class rectipy.observer.Observer(dt: float, record_output: bool = True, record_loss: bool = True, record_vars: list | None = None)[source]

Bases: object

Class that is used to record state variables, outputs, and losses during calls of Network.train, Network.test, or Network.run.

matshow(v: str | Tuple[str, str], ax: Axes | None = None, **kwargs) → Axes[source]

Create a 2D color plot of variable v.

Parameters:

  • v – Tuple that contains the names of the node and the node variable to be plotted.

  • ax – matplotlib.pyplot.Axes instance in which to plot.

  • kwargs – Additional keyword arguments for the matplotlib.pyplot.imshow call.

Returns:

Instance of matplotlib.pyplot.Axes that contains the line plot.

Return type:

plt.Axes

plot(y: str | Tuple[str, str], x: str | Tuple[str, str] | None = None, ax: Axes | None = None, **kwargs) → Axes[source]

Create a line plot with variable y on the y-axis and x on the x-axis.

Parameters:

  • y – Tuple that contains the names of the node and the node variable to be plotted on the y-axis.

  • x – Tuple that contains the names of the node and the node variable to be plotted on the x-axis. If not provided, y will be plotted against time steps.

  • ax – matplotlib.pyplot.Axes instance in which to plot.

  • kwargs – Additional keyword arguments for the matplotlib.pyplot.plot call.

Returns:

Instance of matplotlib.pyplot.Axes that contains the line plot.

Return type:

plt.Axes

record(step: int, output: Tensor, loss: float | Tensor, record_vars: Iterable[Tensor]) → None[source]

Performs a single recording steps.

Parameters:

  • step – Integration step.

  • output – Output of the Network model.

  • loss – Current loss of the Network model.

  • record_vars – Additional variables of the RNN layer that should be recorded.

Return type:

None

property recorded_state_variables: list

RNN state variables that are recorded by this Observer instance.

property recorded_variables: list

RNN state variables that are recorded by this Observer instance.

property recordings
save(key: str, val: Any)[source]

Saves object on observer. Can be retrieved via key.

Parameters:

  • key – Used for storage/retrieval.

  • val – Object to be stored.

to_dataframe(item: str | Tuple[str, str])[source]
to_numpy(item: str | Tuple[str, str]) → ndarray[source]