seed: 0 data: directory: data train_file: piececonst_r421_N1024_smooth1.mat test_file: piececonst_r421_N1024_smooth2.mat train_samples: 1000 test_samples: 200 stride: 5 grid_size: 85 cache: results/darcy85_data.npz model: width: 96 latent_sizes: [512, 256, 128, 64] depths: [3, 1, 1, 1] heads: [4, 4, 8, 8] sinkhorn_iterations: 8 temperature: 1.0 anchor_grad_fraction: 0.25 output_dropout: 0.5 training: epochs: 500 batch_size: 4 learning_rate: 0.001 warmup_epochs: 100 weight_decay: 0.00005 betas: [0.9, 0.99] grad_clip: 1000.0 div_factor: 100 final_div_factor: 100 checkpoint: weight/best_model.pth resume_checkpoint: weight/last_model.pth selection: final_epoch_500_no_test_selection precision: float32 print_every_steps: 25 results: results paper_relative_l2: 0.0070