Training

Trainer, arguments, suites, and pruning:

• TrainingArguments — Training hyperparameters (HF-like)
• Trainer — Generic trainer
• TextPredictionConfig — Config for text prediction
• TextPredictionTrainer — Trainer for text prediction
• TrainerConfig — Shared trainer configuration base
• PrePruneConfig — Pre-pruning
• PostPruneConfig — Post-pruning

Trainer suites

• TrainerSuite — Base suite for paired trainers
• PositiveTrainerSuite — One-vs-rest positive feature suites
• SymmetricTrainerSuite — Symmetric pair suites
• TrainerCollection — Group suites for joint analysis
• SuitePairDefinition — Declarative pair spec
• PositiveFeatureDefinition — Declarative positive-feature spec
• MultiSeedTrainerView — Multi-seed analysis view

Transitions and utilities

• TransitionSpec — Explicit encoder-eval transition (pair(), identity())
• set_seed — Reproducible RNG seeding

Additional symbols (not in the navigation): load_training_stats, create_model_with_gradiend, GradientTrainingDataset, TextGradientTrainingDataset.