Convert a fully qualified class name string to its corresponding class. Args: target: A string representing a class name. Can be eith

add a length column to the dataset df

adds a column with mean

adds a column with mean

creates (weighted) means and stds from mean, std and length cols of the df

Stratifying procedure. Modified from https://vict0rs.ch/2018/05/24/sample-multilabel-dataset/ (based on Sechidis 2011) data is a list of lists: a

Compute ROC curve and ROC area for each class "0"..."n_classes - 1" (or classnames passed via classes), "micro", "macro" returns fpr,tpr,roc (dict

S4 analogue of the pretraining head used in modified CPC https://arxiv.org/abs/2002.02848 (in addition to layer norm) can also be used as global predi

input shape: bs, seq, feat output shape: bs,seq,nc for multi_prediction else bs,nc for global_pool

performs empirical bootstrap https://ocw.mit.edu/courses/mathematics/18-05-introduction-to-probability-and-statistics-spring-2014/readings/MI

S4-compatible load_state_dict with Lightning signature compatibility

Utility method: register a tensor as a buffer or trainable parameter

Utility method: register a tensor as a buffer or trainable parameter

split off a stratified subset of fraction ratio from an original dataset df, stratifies according to a specified label column and age and sex

aggregates potentially multiple predictions per sample (can also pass targs for convenience) idmap: idmap as returned by TimeSeriesCr

Wrap the cuda method to deal with shapes

possibility modify dataset config parameters passed to dataset constructor in derived classes

convolution with padding

Create a schedule with a constant learning rate, using the learning rate set in optimizer. Args: optimizer (:class:`~torch.optim.Opt

Create a schedule with a constant learning rate preceded by a warmup period during which the learning rate increases linearly between 0 and t

Create a schedule with a learning rate that decreases following the values of the cosine function between the initial lr set in the optimizer

Create a schedule with a learning rate that decreases following the values of the cosine function between the initial lr set in the optimizer

Create a schedule with a learning rate that decreases as a with an inverse sqrt law, after a warmup period during which it increases linearl

Create a schedule with a learning rate that decreases linearly from the initial lr set in the optimizer to 0, after a warmup period during wh

Create a schedule with a learning rate that decreases as a polynomial decay from the initial lr set in the optimizer to end lr defined by `lr

Remove the hook from the model.

d_state: the dimension of the state, also denoted by N l_max: the maximum sequence length, also denoted by L if this is not

d_state: the dimension of the state, also denoted by N l_max: the maximum sequence length, also denoted by L if this is not

Calculate (and cache) FFT nodes and their "unprocessed" them with the bilinear transform This should be called everytime the internal length

Calculate (and cache) FFT nodes and their "unprocessed" them with the bilinear transform This should be called everytime the internal length

Version of the step function that has time O(N) instead of O(N^2) per step, which takes advantage of the DPLR form and bilinear discretizatio

Version of the step function that has time O(N) instead of O(N^2) per step, which takes advantage of the DPLR form and bilinear discretizatio

ramp from (e0, t0) -> (e1, t1) through a cosine schedule based on e in [e0, e1]

Model head that takes `nf` features, runs through `lin_ftrs`, and about `nc` classes; added bn and act here

returns submodules assigned to encoder, predictor and head for discriminative learning rates

Function that loads the weights from a given checkpoint file. based on https://github.com/PyTorchLightning/pytorch-lightning/issues/525

override preprocessing in derived classes if desired

metrics: list of metrics to compute ["mae", "r2"]. If None, computes all. target_names: list of names for each regression target

input shape: bs, seq, feat output shape: bs,seq,nc for multi_prediction else bs,nc for global_pool

Construct C~ from C double_length: current C is for length L, convert it to length 2L

Construct C~ from C double_length: current C is for length L, convert it to length 2L

Construct dA and dB for discretized state equation

Construct dA and dB for discretized state equation

Convert string target to class object.

uninterpolated auc as used by sklearn https://github.com/scikit-learn/scikit-learn/blob/1495f6924/sklearn/metrics/ranking.py see also the discussion a

Sequence of batchnorm (if `bn`), dropout (with `p`) and linear (`n_in`,`n_out`) layers followed by `actn`. permute for input of the form B,S

Sequence of batchnorm (if `bn`), dropout (with `p`) and linear (`n_in`,`n_out`) layers followed by `actn`. permute for input of the form B,S