Model Planner Options

Column Processing

• encoder_overrides

Model Architecture

• activation
• aggregation
• channels
• handle_new_target_entities
• module
• normalization
• num_post_message_passing_layers
• num_pre_message_passing_layers
• ranking_embedding_loss_coeff
• output_embedding_dim
• target_embedding_mode
• use_seq_id

Neighbor Sampling

• max_target_neighbors_per_entity

• num_neighbors

Optimization

• base_lr
• batch_size
• early_stopping
• lr_scheduler
• majority_sampling_ratio
• max_epochs
• max_steps_per_epoch
• max_test_steps
• max_val_steps
• weight_decay
• weight_mode

Training Job Plan

• refit_full
• refit_trainval
• run_mode
• metrics
• num_experiments
• tune_metric

Training Table Generation

• entity_candidate_aggregation

• forecast_length

• forecast_type

• lag_timesteps

• split

• timeframe_step

• train_end_offset

• train_start_offset