num_proc: 1 cache_dir: cache/ # Usually for raw and processed data log_dir: logs/ tensorboard_log_dir: tensorboard/ ckpt_dir: ckpt/ run_id: PDSRec # Change this to your customized run id rand_seed: 2024 reproducibility: True max_seq_length: 10 train_batch_size: 256 eval_batch_size: 32 lr: 0.0003 weight_decay: 0.0 warmup_steps: 10000 steps: ~ epochs: 150 max_grad_norm: 1.0 # None for no clipping, else a float value eval_interval: 1 # Evaluate every n epochs patience: 10 # Early stopping. Stop training after n epochs without improvement. Set to None to disable topk: [5,10] metrics: [ndcg,recall] val_metric: recall@5 save: True