| alg_name: "DualEdit" | |
| model_name: "llava-hf/llava-1.5-7b-hf" | |
| device: 0 | |
| # Architecture (from DualEdit paper, LLaVA-1.5 config) | |
| hidden_size: 4096 | |
| adapter_mid_dim: 1024 | |
| cross_att_head_n: 8 | |
| img_tok_n: 576 | |
| # Layer configuration (paper: vision=19, text=16) | |
| vision_adapter_layer: 19 | |
| text_adapter_layer: 16 | |
| llm_layer_tmp: "language_model.model.layers.{}" | |
| # Gating threshold (paper Appendix D: tau=0.6 for LLaVA-1.5) | |
| gating_threshold: 0.3 | |
| # Training (adapted for distributional editing) | |
| edit_lr: 1.e-4 | |
| n_iterations: 5000 | |
| batch_size: 4 | |
| reliability_weight: 1.0 | |
| generality_weight: 1.0 | |
| locality_weight: 1.0 | |
| checkpoint_every: 1000 | |