Quantities
Cockpit offers a large set of so called quantities that can be efficiently tracked during the training process.
|
Loss Quantity class tracking the mini-batch training loss during training. |
|
Parameter Quantitiy class tracking the current parameters in each iteration. |
|
Distance Quantity class tracking distance of the parameters from their init. |
|
Quantity class for tracking parameter update sizes. |
|
Quantitiy Class for tracking the norm of the mean gradient. |
|
Time Quantity Class tracking the time during training. |
|
Alpha Quantity class for the normalized step length α. |
|
CABS Quantity class for the suggested batch size using the CABS criterion. |
|
Quantity class for the evidence-based early-stopping criterion. |
|
Quantity class for one-dimensional histograms of indivdual gradient elements. |
|
Quantity class for two-dimensional histograms over gradient and parameters. |
|
Quantitiy Class for the norm test. |
|
Quantitiy Class for tracking the result of the inner product test. |
|
Quantity Class for the orthogonality test. |
|
Quantity Class tracking the Hessian's largest eigenvalue. |
|
Quantitiy Class tracking the trace of the Hessian during training. |
|
Quantity class for tracking the TIC using diagonal curvature approximation. |
|
Quantity class for the TIC using the trace of curvature and gradient covariance. |
|
Quantitiy Class for the mean gradient signal-to-noise ratio (GSNR). |