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

x: (batch_size, num_leads, seq_len)

x: (batch_size, num_leads, seq_len)

series: (batch_size, num_leads, seq_len) pred: (batch_size, n, patch_size * num_leads) mask: (batch_size, n), 0 is keep, 1 is

series: (batch_size, num_leads, seq_len) pred: (batch_size, num_leads, n, patch_size) mask: (batch_size, num_leads, n), 0 is

returns metrics keys in the order they will later be returned by _eval

Initializes multi-GPU-friendly python logger.

Initializes multi-GPU-friendly python logger.

Computes cosine similary between transfomer pooled output, referred to as input representation, and look-up embedding matrix, that is a dense represen

should return model-specific transforms to bring data from its standard format into the format expected by the foundation model Note: before T

return param (groups), modules=True is only used for freezing

Get the SLURM job ID from environment variables.

Get the next available work directory with incremented version number.

Initialize `m` weights with `func` and set `bias` to 0.

Initialize `m` weights with `func` and set `bias` to 0.

Load and instantiate a model from a full-path YAML config. Args: config_name: Full path to YAML config file (including .yaml ext

Return w, p, q, V, B such that (w - p q^*, B) is unitarily equivalent to the original HiPPO A, B by the matrix V i.e. A = V[w - p q^*]V^*, B

Return w, p, q, V, B such that (w - p q^*, B) is unitarily equivalent to the original HiPPO A, B by the matrix V i.e. A = V[w - p q^*]V^*, B

series: (batch_size, num_leads, seq_len) x: (batch_size, n, patch_size * num_leads)

series: (batch_size, num_leads, seq_len) x: (batch_size, num_leads, n, patch_size)

finetune_dataset e.g. mimic_all_all_all_all_2000_5A -mimic_{subsettrain}_{labelsettrain}_{subsettest}_{labelsettest}_{mincnt}_{digits} where _{di

Perform per-sample random masking by per-sample shuffling. Per-sample shuffling is done by argsort random noise. x: (batch_si

Perform per-sample random masking by per-sample shuffling. Per-sample shuffling is done by argsort random noise. x: (batch_si

Return low-rank matrix L such that A + L is normal

Return low-rank matrix L such that A + L is normal

return the contents of the file named filename or None if file not found

Linear classifier should not change any weights other than the linear layer. This sanity check asserts nothing wrong happens (e.g., BN stats

dataset specific modification in derived classes

evaluate instance-wise scores for a single sample and a single threshold

Step one time step as a recurrent model. Intended to be used during validation. u: (B H) state: (B H N) Returns: output (B H

Step one time step as a recurrent model. Intended to be used during validation. u: (B H) state: (B H N) Returns: output (B H

Store params named in comma-separated `nms` from calling context into attrs in `self`

partitions df into n subsets (with fractions specified by subset_ratios) keeping the label distribution in col_label intact also respecting patie

A, B transition matrices for different measures measure: the type of measure legt - Legendre (translated) legs - Legendre (scaled)

A, B transition matrices for different measures measure: the type of measure legt - Legendre (translated) legs - Legendre (scaled)