| { |
| "model_hash": "e576b74953084f4b29de87857e612650", |
| "model_name": "z_image_dit", |
| "model_class": "diffsynth.models.z_image_dit.ZImageDiT", |
| "quant_config": { |
| "load_prequantized": true, |
| "configs": [ |
| { |
| "method": "entropack_lossy_quant", |
| "backend_config_kwargs": { |
| "target_bpp": 2.193929292929293, |
| "row_rdo_iterations": 2, |
| "row_rdo_candidates": 20 |
| }, |
| "target_modules": [ |
| "noise_refiner.0.attention.to_q", |
| "noise_refiner.0.attention.to_k", |
| "noise_refiner.0.attention.to_v", |
| "noise_refiner.0.feed_forward.w1", |
| "noise_refiner.0.feed_forward.w2", |
| "noise_refiner.0.feed_forward.w3", |
| "noise_refiner.1.attention.to_q", |
| "noise_refiner.1.attention.to_k", |
| "noise_refiner.1.attention.to_v", |
| "noise_refiner.1.feed_forward.w1", |
| "noise_refiner.1.feed_forward.w2", |
| "noise_refiner.1.feed_forward.w3", |
| "context_refiner.0.attention.to_q", |
| "context_refiner.0.attention.to_k", |
| "context_refiner.0.attention.to_v", |
| "context_refiner.0.feed_forward.w1", |
| "context_refiner.0.feed_forward.w2", |
| "context_refiner.0.feed_forward.w3", |
| "context_refiner.1.attention.to_q", |
| "context_refiner.1.attention.to_k", |
| "context_refiner.1.attention.to_v", |
| "context_refiner.1.feed_forward.w1", |
| "context_refiner.1.feed_forward.w2", |
| "context_refiner.1.feed_forward.w3", |
| "layers.0.attention.to_q", |
| "layers.0.attention.to_k", |
| "layers.0.attention.to_v", |
| "layers.0.feed_forward.w1", |
| "layers.0.feed_forward.w2", |
| "layers.0.feed_forward.w3", |
| "layers.1.attention.to_q", |
| "layers.1.attention.to_k", |
| "layers.1.attention.to_v", |
| "layers.1.feed_forward.w1", |
| "layers.1.feed_forward.w2", |
| "layers.1.feed_forward.w3", |
| "layers.2.attention.to_q", |
| "layers.2.attention.to_k", |
| "layers.2.attention.to_v", |
| "layers.2.feed_forward.w1", |
| "layers.2.feed_forward.w2", |
| "layers.2.feed_forward.w3", |
| "layers.3.attention.to_q", |
| "layers.3.attention.to_k", |
| "layers.3.attention.to_v", |
| "layers.3.feed_forward.w1", |
| "layers.3.feed_forward.w2", |
| "layers.3.feed_forward.w3", |
| "layers.4.attention.to_q", |
| "layers.4.attention.to_k", |
| "layers.4.attention.to_v", |
| "layers.4.feed_forward.w1", |
| "layers.4.feed_forward.w2", |
| "layers.4.feed_forward.w3", |
| "layers.5.attention.to_q", |
| "layers.5.attention.to_k", |
| "layers.5.attention.to_v", |
| "layers.5.feed_forward.w1", |
| "layers.5.feed_forward.w2", |
| "layers.5.feed_forward.w3", |
| "layers.6.attention.to_q", |
| "layers.6.attention.to_k", |
| "layers.6.attention.to_v", |
| "layers.6.feed_forward.w1", |
| "layers.6.feed_forward.w2", |
| "layers.6.feed_forward.w3", |
| "layers.7.attention.to_q", |
| "layers.7.attention.to_k", |
| "layers.7.attention.to_v", |
| "layers.7.feed_forward.w1", |
| "layers.7.feed_forward.w2", |
| "layers.7.feed_forward.w3", |
| "layers.8.attention.to_q", |
| "layers.8.attention.to_k", |
| "layers.8.attention.to_v", |
| "layers.8.feed_forward.w1", |
| "layers.8.feed_forward.w2", |
| "layers.8.feed_forward.w3", |
| "layers.9.attention.to_q", |
| "layers.9.attention.to_k", |
| "layers.9.attention.to_v", |
| "layers.9.feed_forward.w1", |
| "layers.9.feed_forward.w2", |
| "layers.9.feed_forward.w3", |
| "layers.10.attention.to_q", |
| "layers.10.attention.to_k", |
| "layers.10.attention.to_v", |
| "layers.10.feed_forward.w1", |
| "layers.10.feed_forward.w2", |
| "layers.10.feed_forward.w3", |
| "layers.11.attention.to_q", |
| "layers.11.attention.to_k", |
| "layers.11.attention.to_v", |
| "layers.11.feed_forward.w1", |
| "layers.11.feed_forward.w2", |
| "layers.11.feed_forward.w3", |
| "layers.12.attention.to_q", |
| "layers.12.attention.to_k", |
| "layers.12.attention.to_v", |
| "layers.12.feed_forward.w1", |
| "layers.12.feed_forward.w2", |
| "layers.12.feed_forward.w3", |
| "layers.13.attention.to_q", |
| "layers.13.attention.to_k", |
| "layers.13.attention.to_v", |
| "layers.13.feed_forward.w1", |
| "layers.13.feed_forward.w2", |
| "layers.13.feed_forward.w3", |
| "layers.14.attention.to_q", |
| "layers.14.attention.to_k", |
| "layers.14.attention.to_v", |
| "layers.14.feed_forward.w1", |
| "layers.14.feed_forward.w2", |
| "layers.14.feed_forward.w3", |
| "layers.15.attention.to_q", |
| "layers.15.attention.to_k", |
| "layers.15.attention.to_v", |
| "layers.15.feed_forward.w1", |
| "layers.15.feed_forward.w2", |
| "layers.15.feed_forward.w3", |
| "layers.16.attention.to_q", |
| "layers.16.attention.to_k", |
| "layers.16.attention.to_v", |
| "layers.16.feed_forward.w1", |
| "layers.16.feed_forward.w2", |
| "layers.16.feed_forward.w3", |
| "layers.17.attention.to_q", |
| "layers.17.attention.to_k", |
| "layers.17.attention.to_v", |
| "layers.17.feed_forward.w1", |
| "layers.17.feed_forward.w2", |
| "layers.17.feed_forward.w3", |
| "layers.18.attention.to_q", |
| "layers.18.attention.to_k", |
| "layers.18.attention.to_v", |
| "layers.18.feed_forward.w1", |
| "layers.18.feed_forward.w2", |
| "layers.18.feed_forward.w3", |
| "layers.19.attention.to_q", |
| "layers.19.attention.to_k", |
| "layers.19.attention.to_v", |
| "layers.19.feed_forward.w1", |
| "layers.19.feed_forward.w2", |
| "layers.19.feed_forward.w3", |
| "layers.20.attention.to_q", |
| "layers.20.attention.to_k", |
| "layers.20.attention.to_v", |
| "layers.20.feed_forward.w1", |
| "layers.20.feed_forward.w2", |
| "layers.20.feed_forward.w3", |
| "layers.21.attention.to_q", |
| "layers.21.attention.to_k", |
| "layers.21.attention.to_v", |
| "layers.21.feed_forward.w1", |
| "layers.21.feed_forward.w2", |
| "layers.21.feed_forward.w3", |
| "layers.22.attention.to_q", |
| "layers.22.attention.to_k", |
| "layers.22.attention.to_v", |
| "layers.22.feed_forward.w1", |
| "layers.22.feed_forward.w2", |
| "layers.22.feed_forward.w3", |
| "layers.23.attention.to_q", |
| "layers.23.attention.to_k", |
| "layers.23.attention.to_v", |
| "layers.23.feed_forward.w1", |
| "layers.23.feed_forward.w2", |
| "layers.23.feed_forward.w3", |
| "layers.24.attention.to_q", |
| "layers.24.attention.to_k", |
| "layers.24.attention.to_v", |
| "layers.24.feed_forward.w1", |
| "layers.24.feed_forward.w2", |
| "layers.24.feed_forward.w3", |
| "layers.25.attention.to_q", |
| "layers.25.attention.to_k", |
| "layers.25.attention.to_v", |
| "layers.25.feed_forward.w1", |
| "layers.25.feed_forward.w2", |
| "layers.25.feed_forward.w3", |
| "layers.26.attention.to_q", |
| "layers.26.attention.to_k", |
| "layers.26.attention.to_v", |
| "layers.26.feed_forward.w1", |
| "layers.26.feed_forward.w2", |
| "layers.26.feed_forward.w3", |
| "layers.27.attention.to_q", |
| "layers.27.attention.to_k", |
| "layers.27.attention.to_v", |
| "layers.27.feed_forward.w1", |
| "layers.27.feed_forward.w2", |
| "layers.27.feed_forward.w3", |
| "layers.28.attention.to_q", |
| "layers.28.attention.to_k", |
| "layers.28.attention.to_v", |
| "layers.28.feed_forward.w1", |
| "layers.28.feed_forward.w2", |
| "layers.28.feed_forward.w3", |
| "layers.29.attention.to_q", |
| "layers.29.attention.to_k", |
| "layers.29.attention.to_v", |
| "layers.29.feed_forward.w1", |
| "layers.29.feed_forward.w2", |
| "layers.29.feed_forward.w3" |
| ] |
| }, |
| { |
| "method": "entropack_lossy_quant", |
| "backend_config_kwargs": { |
| "target_bpp": 3.0, |
| "row_rdo_iterations": 2, |
| "row_rdo_candidates": 20 |
| }, |
| "target_modules": [ |
| "noise_refiner.0.attention.to_out.0", |
| "noise_refiner.1.attention.to_out.0", |
| "context_refiner.0.attention.to_out.0", |
| "context_refiner.1.attention.to_out.0", |
| "layers.0.attention.to_out.0", |
| "layers.1.attention.to_out.0", |
| "layers.2.attention.to_out.0", |
| "layers.3.attention.to_out.0", |
| "layers.4.attention.to_out.0", |
| "layers.5.attention.to_out.0", |
| "layers.6.attention.to_out.0", |
| "layers.7.attention.to_out.0", |
| "layers.8.attention.to_out.0", |
| "layers.9.attention.to_out.0", |
| "layers.10.attention.to_out.0", |
| "layers.11.attention.to_out.0", |
| "layers.12.attention.to_out.0", |
| "layers.13.attention.to_out.0", |
| "layers.14.attention.to_out.0", |
| "layers.15.attention.to_out.0", |
| "layers.16.attention.to_out.0", |
| "layers.17.attention.to_out.0", |
| "layers.18.attention.to_out.0", |
| "layers.19.attention.to_out.0", |
| "layers.20.attention.to_out.0", |
| "layers.21.attention.to_out.0", |
| "layers.22.attention.to_out.0", |
| "layers.23.attention.to_out.0", |
| "layers.24.attention.to_out.0", |
| "layers.25.attention.to_out.0", |
| "layers.26.attention.to_out.0", |
| "layers.27.attention.to_out.0", |
| "layers.28.attention.to_out.0", |
| "layers.29.attention.to_out.0" |
| ] |
| }, |
| { |
| "method": "entropack_lossy_quant", |
| "backend_config_kwargs": { |
| "target_bpp": 4.0, |
| "row_rdo_iterations": 2, |
| "row_rdo_candidates": 20 |
| }, |
| "target_modules": [ |
| "all_x_embedder.2-1", |
| "all_final_layer.2-1.linear", |
| "all_final_layer.2-1.adaLN_modulation.1", |
| "noise_refiner.0.adaLN_modulation.0", |
| "noise_refiner.1.adaLN_modulation.0", |
| "t_embedder.mlp.0", |
| "t_embedder.mlp.2", |
| "cap_embedder.1", |
| "layers.0.adaLN_modulation.0", |
| "layers.1.adaLN_modulation.0", |
| "layers.2.adaLN_modulation.0", |
| "layers.3.adaLN_modulation.0", |
| "layers.4.adaLN_modulation.0", |
| "layers.5.adaLN_modulation.0", |
| "layers.6.adaLN_modulation.0", |
| "layers.7.adaLN_modulation.0", |
| "layers.8.adaLN_modulation.0", |
| "layers.9.adaLN_modulation.0", |
| "layers.10.adaLN_modulation.0", |
| "layers.11.adaLN_modulation.0", |
| "layers.12.adaLN_modulation.0", |
| "layers.13.adaLN_modulation.0", |
| "layers.14.adaLN_modulation.0", |
| "layers.15.adaLN_modulation.0", |
| "layers.16.adaLN_modulation.0", |
| "layers.17.adaLN_modulation.0", |
| "layers.18.adaLN_modulation.0", |
| "layers.19.adaLN_modulation.0", |
| "layers.20.adaLN_modulation.0", |
| "layers.21.adaLN_modulation.0", |
| "layers.22.adaLN_modulation.0", |
| "layers.23.adaLN_modulation.0", |
| "layers.24.adaLN_modulation.0", |
| "layers.25.adaLN_modulation.0", |
| "layers.26.adaLN_modulation.0", |
| "layers.27.adaLN_modulation.0", |
| "layers.28.adaLN_modulation.0", |
| "layers.29.adaLN_modulation.0" |
| ] |
| } |
| ] |
| } |
| } |
