Human-readable string representation for display.

Human-readable string representation for display.

Sequence of allowed chars, order respected.

Sequence of allowed keys, order respected.

Sequence of allowed unicode values, order respected.

Collates a list of samples into a padded batch of inputs and targets. Each input sample in the list should be a tuple of (input, target) tenso

Number of CPUs to request per task per node taking into account the number of GPUs and dataloading workers.

Sequence of decoded labels in the path leading to this beam state, ignoring the blank label at the trie root.

The duration of the EMG session in minutes.

The names of the fields in the ``timeseries`` compound HDF5 Dataset.

Sequence of keys recorded by the keylogger during the data-collection session along with the press and release timestamps for each key

Last label corresponding to this decoding state.

Convert a single label to its corresponding textual representation after standardizing.

Fetch the unicode value for the given categorical label.

Integer labels corresponding to the label string.

LM score corresponding to this beam node.

Sequence of LM scores in the path leading to this beam state.

LM state corresponding to this beam state.

Sequence of LM states in the path leading to this beam state.

Categorical label of the null-class (blank label).

Number of training classes including null-class (blank label).

Total probability (in log-space) of the decoding path leading to this state.

Sequence of sentences prompted to the user during the data-collection session along with the start and end timestamps for each prompt.

Sample `n` users from the given dataset who have at least `min_sessions` sessions.

Unique name of the session.

Load and return a contiguous slice of the timeseries windowed by the provided start and end timestamps. Args: start_t (fl

Test BeamState with LM.

Test `BeamState` and `TrieNode` operations when several paths lead to same decoding. Two paths leading to the same decoding prefix should mat

Test BeamState without LM.

Test CTCGreedyDecoder basics.

Test CTCGreedyDecoder.decode_batch() API for batched decoding.

Test CTCBeamDecoder delete handling. Total lm score for a word should be the same irrespective of deletes.

Test that the decoder tracks onset timestamps correctly despite several paths leading to the same decoding. Consider the following two paths

Last timestamp corresponding to this state.

Sequence of onset timestamps corresponding to the decoded labels in the path leading to this beam state.

EMG timestamps. NOTE: This reads the entire sequence of timesetamps from the underlying HDF5 file and therefore incurs disk latency.

Fetch the key corresponding to the given unicode value.

Fetch the categorical label for the given unicode value.

Unique ID of the user this session corresponds to.