Python API catalog

Search classes and functions by name or module. Documentation is generated from the Python source without importing training dependencies.

216 of 216 APIs

Search Python API

Name	Description
analyze	Exclusively write local diagnostics, retaining series IDs only in the file.
AnnotatedDataset	Small source adapter; callers retain control of training and orchestration.
ApneaTask	Sleep Apnea Task
Artifact	`dataclass`
assign	Hash-rank all groups, independent of input ordering, labels, and model scores.
audit	Inspect local source labels without opening files for writing or training a model.
audit	Count candidate coverage using unchanged originals bound to trusted local evidence.
authenticate_nsrr	Authenticate the [NSRR](https://sleepdata.org) download token.
BasicCodec	Version of this codec.
build_candidates	Build candidates from one series' raw CSV groups and independently verified clock.
build_model
build_model	Build the Keras model while using heliaEDGE's registered PatchLayer2D.
build_model
CandidateLabels	`dataclass`
Check	`dataclass`
class_metrics	Fixed two-class descriptive metrics; no probability interpretation.
ClientInfo
ClientManager
CmidssDataset	CMIDSS dataset
CmidssSleepStage	CMIDSS sleep stages
Codec
CodecError
collect_calibration	Collect frozen-normalizer model inputs from native train contexts.
command
command
compare	Run one declared descriptive comparison; output is exclusively local evidence.
compare_labels
compute_apnea_efficiency	Compute apnea efficiency.
compute_apnea_hypopnea_index	Compute apnea hypopnea index (AHI).
compute_sleep_apnea_durations	Compute sleep apnea durations
compute_sleep_efficiency	Compute sleep efficiency.
compute_sleep_stage_durations	Compute sleep stage durations
compute_total_sleep_time	Compute total sleep time (# samples).
Config	`dataclass`
Config	`dataclass`
Config	`dataclass`
ConnectionClosed
context_counts	Count native complete contexts without compacting invalid or unknown frames.
contexts	Yield complete nonoverlapping model contexts, never bridge dropped frames.
Crc16
create_augmentation_layer	Create an augmentation layer from a configuration
create_augmentation_pipeline	Create an augmentation pipeline from a list of augmentation configurations.
create_data_pipeline	Create a data pipeline.
create_data_pipeline	Create a data pipeline.
create_reference	Record deterministic synthetic I/O. This establishes no task accuracy.
create_split	Split an explicit cohort deterministically, before reading features or labels.
dataBlock
dataBlock
dataset
Dataset	Dataset serves as a base class to download and provide unified access to datasets.
dataType
dataType
declare_golden	Freeze prospective identities/configuration without training or reading labels.
demo	Sleep Apnea Demo
demo	Run sleep stage classification demo.
DetectionRuntime	One context in, float32 logits out; never applies a softmax.
download_bundle	Download an explicitly selected revision and validate its artifact inventory.
download_file	Download file from supplied url to destination streaming.
download_nsrr	Recursively download files from [NSRR](https://sleepdata.org).
download_nsrr_file	Download a file from `url` to `dst` and verify `checksum`.
environment
evaluate
evaluate	Evaluate one split with the shared unweighted classification accumulator.
evaluate	Write an aggregate report only after evaluation and input revalidation.
evaluate	Pooled scored-epoch metrics; unlike Keras history, not averaged batch means.
evaluate	Evaluate sleep apnea model.
evaluate	Evaluate sleep stage model.
evaluate_quantized	Replay and match the frozen test index; conversion is already fixed.
evaluate_windows	Apply the scoring mask after inference so excluded epochs retain context.
evb_to_pcClient
evb_to_pcService
EvbBackend	EVB inference engine backend
examples
export	Export sleep apnea model.
export	Export sleep stage model.
export_bundle
extract	Read aligned samples, using an optional independent source UTC clock for cadence.
FeatureParams	Feature configuration params
Features	`dataclass`
FeatureSet	Feature set abstract class.
filename
fingerprint
fingerprint
FramedTransport
freeze	Freeze all audited subjects; coverage summarizes an already assigned split.
FS_C_EAR_9	Feature set: FS-C-EAR-9
FS_H_E_10	Feature set: FS-H-E-10
FS_W_A_5	Feature set: FS-W-A-5
FS_W_P_40	Feature set: FS-W-P-40
FS_W_P_5	Feature set: FS-W-P-5
FS_W_PA_14	Feature set: FS-W-PA-14
get_apnea_color_map	Get color map for apnea classes
get_serial_transport	Create serial transport. Scans looking for port matching criteria.
get_stage_color_map	Get color map for sleep stages
H5Dataloader	Dataloader to load features from folder containing HDF5 files.
Ievb_to_pc
InferenceBackend	Backend inference engine base class
inspect_clock	Report local-clock discontinuities; do not infer physical sample gaps.
inspect_integer_graph	Inspect every FlatBuffer subgraph tensor; integer I/O alone is insufficient.
inspect_nights	Return unambiguous paired sleep candidates in source-step coordinates.
Ipc_to_evb
LIBUSBSIOI2CTransport
LIBUSBSIOSPITransport
list_nsrr_items	Recursively list items for a dataset to be downloaded.
load_events	Retain every row so missing/duplicate event pairs cannot disappear silently.
load_split
load_split
main
main
main
main
main
main
main
main
main
main
main
make_data	Create deterministic train/validation/test raw waveforms and integer targets.
make_fixture
match_context	Reject coordinate or target drift even when shapes and timestamps look plausible.
materialize	Return provenance after atomically creating a new clock-bearing source folder.
MesaDataset	MESA dataset
MesaSleepStage	MESA sleep stages
MessageType
ModelFactoryItem
NamedParams	Named parameters is used to store parameters for a specific model, preprocessing, or augmentation.
Normalizer	`dataclass`
output_contract
package	Save model, metrics, config, and reload vectors into an archive-profile bundle.
parse_content	Parse file or raw content into Pydantic model.
pc_to_evbClient
pc_to_evbService
PcBackend	PC inference engine backend.
PlotPallette	`dataclass`
predict
prepare	Cache stateless features only: annotations and normalization never enter the key.
prepare	Convert finite [examples, 128] waveforms into deterministic [examples, 128, 1] features.
prepare_partition	Preserve each record's context; omit only windows with no scored epochs.
prepare_subject	Copy, map labels, impute invalid rows, and normalize the complete record.
PreparedPartition	`dataclass`
PreparedRecord	`dataclass`
profile_membership	Profile training only, keeping the reader's integrity checks and finite schedule.
publish_bundle	Dry run by default. Upload a validated snapshot without loading a model.
QuantizationParams	Quantization parameters
quantize_run	One fixed train-only calibration, conversion and evaluation; no tuning loop.
read_bound_json	Parse the same bytes whose digest was verified, including transient changes.
read_cohort	Read an explicit JSON list of unique filename stems, independent of discovery order.
read_json
read_recording
read_subject	Legacy two-array interface; use read_recording to retain the independent clock.
read_subject	Read three self-contained HDF5 datasets and validate their schema.
read_verified_features	Read and verify legacy features, then apply whole-record normalization.
reconstruct_legacy_features	Reconstruct FS-W-A-5 features without reading historical labels.
Recording	`dataclass`
Reference	Simple container class used for pass by reference.
replay_bundle	Replay every declared model against its recorded, hash-bound synthetic I/O.
RequestContext
RequestError
require_self_contained	Reject indirect storage while allowing ordinary internal hard links.
revision	Best-effort checkout identity; implementation hashes remain authoritative.
RpcCommands	RPC commands
RpmsgTransport
run	sleepKIT CLI
run	Historical-label recipe, preserved for existing experiments.
run	Generate, train, evaluate, and package one synthetic recipe run.
run	Declare, prepare, train, evaluate and archive one fixed-model experiment.
run_membership	Compose an AnnotatedDataset with training, evaluation, and versioned export.
run_smoke
SD2Runtime	Lazy LiteRT runtime pinned to the reviewed SD-2-TCN-SM model bytes.
SerialTransport
Server
ServerThread
Service
setup_plotting	Setup plotting environment for matplotlib and plotly
SimpleServer
SleepApnea	Sleep apnea class
SleepStage	Sleep stage class
stage_baseline	Package a pinned historical TFLite baseline without conversion or retraining.
stage_bundle	Copy an explicit artifact list into an immutable directory atomically.
stage_release	Preserve model bytes/evidence and add caller-selected release documentation.
StagesDataset	STAGES dataset
StagesSleepStage	Sleep stage enum
StageTask	Sleep Stage Task
status
status
subject_data_preprocessor	Preprocess entire subject data.
subject_data_preprocessor	Preprocess entire subject data.
subject_features
SubjectRecord	`dataclass`
summarize	Verify local run evidence and exclusively write a per-series JSON report.
Task	Task base class. All tasks should inherit from this class.
TaskMode	Task run mode
TaskParams	Task configuration params
TCPTransport
TensorSpec	`dataclass`
train	Train finite epochs with retained tail batches and return ``(model, history)``.
train	Fit every supervised context once per epoch using Keras array adapters.
train	Train sleep apnea model.
train	Train sleep stage model.
training_model	Build and compile the recipe's logits model; usable without running the pipeline.
Transport	Base transport class.
TransportArbitrator	Shares a transport between a server and multiple clients.
validate_bundle	Verify structure/integrity and optional persisted runtime evidence.
validate_input	Validate source clocks before any cache access; clocks are never inferred from row indices.
validate_model	Require the saved SS-3-TCN-SM float32 sequence-to-logits interface.
validate_output
validate_split	Return a normalized copy; filename stems are caller-declared subject identities.
verified_calibration	Use the exact persisted calibration array whose hash is declared publicly.
verify	Read all samples and optionally check event clocks, not annotation semantics.
window_subject	Return nonoverlapping 240-epoch windows without stitching gaps or records.
WindowedRecord	`dataclass`
write_index	Write JSONL in evaluation order without retaining a dataset-wide row list.
write_json
YsywDataset	YSYW dataset
YsywSleepStage	YSYW sleep stages