Datasets:
Modalities:
Text
Formats:
json
Languages:
English
Size:
< 1K
Tags:
reinforcement-learning
policy-optimization
klpo
exact-moment-replay
score-centering
stratified-sampling
License:
| """Celestis-RL: a standalone research implementation, not pretrained model weights.""" | |
| from .estimators import Correction, draw_correction, exact_correction, plan_head | |
| from .losses import token_loss, selected_token_loss, sequence_loss | |
| from .scoring import selected_linear_scores | |
| from .certificates import ConfidenceLedger, AuditResult, paired_gain_audit | |
| from .transaction import transactional_step, StepResult | |
| __version__ = "2.0.0" | |
| __all__ = ["Correction", "draw_correction", "exact_correction", "plan_head", | |
| "token_loss", "selected_token_loss", "sequence_loss", "selected_linear_scores", | |
| "ConfidenceLedger", "AuditResult", "paired_gain_audit", "transactional_step", "StepResult"] | |
| from .stratified import stratified_correction, coalesce_correction | |
| from .streaming import linear_token_loss, centered_linear_scores | |
| from .moment_replay import FeatureBank, MomentBatch, capture_moments, moment_token_loss, moment_sequence_loss, moment_dimension | |
| from .sequential_audit import MetricSpec, SequentialParetoAudit | |
| __all__ += ["stratified_correction", "coalesce_correction", "linear_token_loss", "centered_linear_scores", | |
| "FeatureBank", "MomentBatch", "capture_moments", "moment_token_loss", "moment_sequence_loss", | |
| "MetricSpec", "SequentialParetoAudit"] | |
| from .transport import AffineBankTransport | |
| __all__ += ["AffineBankTransport", "moment_dimension"] | |