protosleepnet.baselines.test_seq_1ch

Evaluate single-channel SeqSleepNet-Phan on MASS (in-domain).

Loads the pretrained model from HuggingFace and evaluates on the MASS test split using per-subject voting (L=20). Saves per-subject predictions and aggregated metrics.

Usage:

python -m protosleepnet.baselines.test_seq_1ch –gpu_id 0

Attributes

CHANNELS
CLASS_NAMES
MODEL_NAME
PIPELINE
SEQ_LEN

Functions

compute_metrics(all_proba, all_targets[, ignore_index])	Compute metrics from aggregated probabilities and targets.
evaluate_subject(model, inputs, L, device)	Sliding-window voting evaluation for a single subject.
main()

Module Contents

protosleepnet.baselines.test_seq_1ch.compute_metrics(all_proba, all_targets, ignore_index=-1)

Compute metrics from aggregated probabilities and targets.

protosleepnet.baselines.test_seq_1ch.evaluate_subject(model, inputs, L, device)

Sliding-window voting evaluation for a single subject.

protosleepnet.baselines.test_seq_1ch.main()

protosleepnet.baselines.test_seq_1ch.CHANNELS = ['EEG']

protosleepnet.baselines.test_seq_1ch.CLASS_NAMES = ['W', 'N1', 'N2', 'N3', 'REM']

protosleepnet.baselines.test_seq_1ch.MODEL_NAME = 'seqsleepnet-phan'

protosleepnet.baselines.test_seq_1ch.PIPELINE = 'seqsleepnet'

protosleepnet.baselines.test_seq_1ch.SEQ_LEN = 20