Core (Machine Gnostics MAGNET)¶
The machinegnostics.magnet module provides the fundamental infrastructure for MAGNET: the Tensor facade, History, Callback, EarlyStopping, runtime configuration, and gnostic characteristic computation.
Key Core Classes¶
- Tensor — Torch-backed tensor facade with
data,grad,shape, arithmetic,backward(),zero_grad() - History — Training history container (
record(),last()) - Callback — Base training callback (
set_model,on_train_*,on_epoch_*) - EarlyStopping — Stop training when monitored metric stalls (
monitor='val_loss',patience=5)
Runtime Configuration¶
configure(device='auto', dtype='float64', seed=None, deterministic=True)get_runtime(),get_torch_device(),get_torch_dtype()to_numpy(),to_torch()— Conversion helpers
Gnostic Helpers¶
compute_characteristics()— Computefi,fj,hi,hjfrom valuescustom_tensor()— Bridge NumPy results into torch autogradgnostic_weights_i()/gnostic_weights_j()— Weight computations
Author: Nirmal Parmar