{ "weight_decay": 0.01, "max_lr": 8e-5, "min_lr": 1e-5, "warmup_steps": 10000, "d_model": 1024, "nhead": 16, "num_layers": 24, "dropout": 0.0, "patch_size": 2, "in_channels": 4, "sigma_emb_hz": 10000, "rope_hz": 10000, "batch_size": 12, "target_size_H": 512, "target_size_W": 512, "text_dim": 1024, "max_text_len": 77, "drop_prob": 0.1 }