quantdata.dysanos.grid.rndysanos.conditional_tail_experiment#
Full-universe training and one-sided-state analytics for conditional tails.
- class quantdata.dysanos.grid.rndysanos.conditional_tail_experiment.ConditionalTailTracker(epochs=<factory>, batch_loss=<factory>, learning_rate=<factory>, gradient_norm=<factory>, valuation_epochs=<factory>, training_values=<factory>, validation_values=<factory>, restart_epochs=<factory>, converged_epoch=None)[source]#
Bases:
objectCompact restartable optimization and full-valuation history.
- class quantdata.dysanos.grid.rndysanos.conditional_tail_experiment.ConditionalTailTrainConfig(*, batch_size=512, valuation_batch_size=2048, lr=0.001, final_lr=1e-06, weight_decay=0.0001, compile=True, value_update=5000, checkpoint_update=10000, minimum_epochs=50000, convergence_points=5, convergence_relative_tolerance=0.002, convergence_log_coverage_tolerance=0.1, checkpoint_root='~/.cache/rndysanos/conditional-tail')[source]#
Bases:
objectOperational configuration for conditional-tail optimization.
- class quantdata.dysanos.grid.rndysanos.conditional_tail_experiment.ConditionalTailTrainingResult(model, training, validation, tracker, checkpoint_path, restarted)[source]#
Bases:
objectTrained model, complete final valuations, and checkpoint metadata.
- quantdata.dysanos.grid.rndysanos.conditional_tail_experiment.prepare_raw_option_tail_data(universe, *, env)[source]#
Prepare state, raw option price changes, values, and roundoff bands.
- quantdata.dysanos.grid.rndysanos.conditional_tail_experiment.process_memory(device)[source]#
Return process and CUDA peak memory in GiB.
- quantdata.dysanos.grid.rndysanos.conditional_tail_experiment.split_tail_data(data, *, validation_fraction)[source]#
Split paths contiguously so validation never enters normalization/training.