{ "_class_name": "ZImageTransformer2DModel", "_diffusers_version": "0.36.0.dev0", "all_f_patch_size": [ 1 ], "all_patch_size": [ 2 ], "axes_dims": [ 32, 48, 48 ], "axes_lens": [ 1536, 512, 512 ], "cap_feat_dim": 2560, "dim": 3840, "in_channels": 16, "n_heads": 30, "n_kv_heads": 30, "n_layers": 30, "n_refiner_layers": 2, "norm_eps": 1e-05, "qk_norm": true, "rope_theta": 256.0, "t_scale": 1000.0, "quantization_config": { "activation_scheme": "dynamic", "fmt": "e4m3", "quant_method": "fp8", "weight_block_size": [ 128, 128 ], "modules_to_not_convert": [ "layers.15.adaLN_modulation.0", "layers.13.attention.norm_k", "noise_refiner.1.feed_forward.w1", "layers.29.feed_forward.w2", "layers.3.attention.norm_q", "layers.27.ffn_norm2", "layers.11.feed_forward.w3", "layers.1.feed_forward.w3", "layers.26.feed_forward.w2", "noise_refiner.0.attention_norm2", "layers.7.adaLN_modulation.0", "layers.17.adaLN_modulation.0.bias", "noise_refiner.1.ffn_norm1", "layers.11.attention.norm_k", "context_refiner.0.ffn_norm2", "layers.22.attention_norm2", "layers.6.adaLN_modulation.0", "layers.20.attention_norm1", "layers.29.feed_forward.w1", "layers.12.attention.norm_k", "layers.28.feed_forward.w1", "layers.22.adaLN_modulation.0.bias", "layers.20.adaLN_modulation.0", "noise_refiner.1.adaLN_modulation.0", "layers.21.attention.norm_q", "layers.4.attention.norm_q", "layers.28.attention_norm1", "layers.22.attention.norm_k", "cap_pad_token", "t_embedder.mlp.2.bias", "layers.21.attention_norm1", "layers.20.feed_forward.w2", "layers.24.adaLN_modulation.0", "layers.11.adaLN_modulation.0", "layers.16.feed_forward.w2", "layers.21.ffn_norm1", "layers.8.feed_forward.w2", "layers.3.feed_forward.w2", "layers.28.ffn_norm1", "layers.11.attention.norm_q", "layers.27.adaLN_modulation.0", "noise_refiner.1.attention.norm_q", "layers.6.feed_forward.w2", "layers.15.attention.norm_k", "layers.17.attention_norm1", "layers.28.adaLN_modulation.0", "layers.21.attention_norm2", "layers.13.attention.norm_q", "all_x_embedder.2-1.bias", "cap_embedder.0", "layers.4.attention.norm_k", "layers.4.feed_forward.w1", "context_refiner.1.ffn_norm2", "layers.29.attention_norm1", "layers.3.adaLN_modulation.0.bias", "layers.24.feed_forward.w2", "layers.12.feed_forward.w2", "layers.5.adaLN_modulation.0", "layers.27.attention_norm2", "layers.9.ffn_norm1", "layers.4.ffn_norm2", "x_pad_token", "layers.20.feed_forward.w1", "layers.26.ffn_norm1", "layers.16.ffn_norm1", "layers.10.attention.norm_q", "layers.20.attention_norm2", "context_refiner.0.attention_norm1", "layers.17.feed_forward.w2", "layers.27.attention.norm_q", "layers.22.feed_forward.w2", "layers.25.ffn_norm1", "layers.6.ffn_norm1", "layers.2.feed_forward.w2", "layers.27.feed_forward.w3", "layers.15.feed_forward.w1", "layers.19.attention.norm_q", "layers.6.attention.norm_k", "layers.18.adaLN_modulation.0.bias", "layers.28.feed_forward.w3", "layers.10.attention.norm_k", "layers.8.attention.norm_k", "layers.3.feed_forward.w3", "layers.23.attention.norm_q", "t_embedder.mlp.2", "layers.29.attention.norm_q", "layers.14.adaLN_modulation.0.bias", "noise_refiner.0.ffn_norm2", "layers.17.feed_forward.w1", "cap_embedder.1.bias", "noise_refiner.0.feed_forward.w3", "layers.10.adaLN_modulation.0", "layers.27.attention_norm1", "layers.6.attention.norm_q", "layers.5.attention.norm_q", "layers.29.feed_forward.w3", "layers.29.adaLN_modulation.0.bias", "layers.23.ffn_norm2", "layers.15.attention.norm_q", "context_refiner.1.feed_forward.w1", "layers.2.attention.norm_k", "layers.20.feed_forward.w3", "noise_refiner.0.feed_forward.w1", "layers.25.feed_forward.w1", "layers.25.attention.norm_k", "layers.8.feed_forward.w3", "noise_refiner.0.attention.norm_q", "layers.2.adaLN_modulation.0", "noise_refiner.1.adaLN_modulation.0.bias", "layers.22.ffn_norm2", "layers.5.ffn_norm1", "layers.28.ffn_norm2", "layers.5.attention_norm2", "all_final_layer.2-1.adaLN_modulation.1.bias", "layers.9.adaLN_modulation.0", "layers.14.feed_forward.w3", "layers.13.feed_forward.w3", "layers.5.attention_norm1", "layers.23.adaLN_modulation.0", "layers.13.attention_norm1", "layers.21.ffn_norm2", "layers.18.ffn_norm2", "context_refiner.0.feed_forward.w1", "cap_embedder.1", "layers.8.adaLN_modulation.0", "layers.23.adaLN_modulation.0.bias", "layers.8.attention_norm1", "layers.15.adaLN_modulation.0.bias", "noise_refiner.1.attention.norm_k", "all_final_layer.2-1.adaLN_modulation.1", "layers.21.feed_forward.w2", "noise_refiner.0.attention_norm1", "layers.13.ffn_norm2", "layers.15.feed_forward.w3", "layers.1.attention.norm_q", "layers.4.attention_norm1", "layers.12.attention_norm1", "layers.17.feed_forward.w3", "layers.17.adaLN_modulation.0", "layers.7.feed_forward.w1", "layers.6.attention_norm2", "layers.14.feed_forward.w1", "layers.10.adaLN_modulation.0.bias", "layers.26.adaLN_modulation.0.bias", "layers.5.feed_forward.w2", "layers.10.feed_forward.w1", "layers.15.ffn_norm1", "layers.15.attention_norm1", "layers.11.feed_forward.w1", "layers.21.adaLN_modulation.0.bias", "layers.24.attention_norm1", "layers.0.adaLN_modulation.0", "layers.16.attention_norm2", "context_refiner.1.ffn_norm1", "layers.0.attention.norm_k", "layers.9.attention_norm1", "layers.12.adaLN_modulation.0", "layers.12.attention_norm2", "layers.11.ffn_norm2", "layers.23.attention.norm_k", "layers.21.feed_forward.w1", "layers.18.adaLN_modulation.0", "layers.26.attention.norm_k", "layers.0.feed_forward.w2", "layers.28.attention_norm2", "layers.10.feed_forward.w3", "all_final_layer.2-1.linear", "noise_refiner.0.adaLN_modulation.0.bias", "layers.29.attention.norm_k", "layers.26.feed_forward.w1", "layers.24.feed_forward.w3", "noise_refiner.0.ffn_norm1", "layers.1.attention_norm2", "layers.14.attention_norm1", "layers.19.attention_norm2", "layers.3.ffn_norm2", "layers.26.ffn_norm2", "layers.19.feed_forward.w2", "layers.9.feed_forward.w1", "all_x_embedder.2-1", "layers.0.attention_norm2", "layers.29.ffn_norm2", "layers.0.attention.norm_q", "layers.16.ffn_norm2", "layers.3.attention.norm_k", "layers.27.attention.norm_k", "layers.28.attention.norm_k", "noise_refiner.1.ffn_norm2", "layers.18.attention.norm_q", "layers.0.adaLN_modulation.0.bias", "layers.0.ffn_norm1", "layers.1.feed_forward.w1", "layers.25.feed_forward.w2", "layers.5.feed_forward.w1", "layers.1.adaLN_modulation.0.bias", "layers.18.ffn_norm1", "layers.20.ffn_norm1", "context_refiner.0.feed_forward.w2", "layers.24.ffn_norm1", "context_refiner.1.attention_norm1", "layers.20.attention.norm_k", "layers.7.ffn_norm1", "layers.27.ffn_norm1", "layers.7.attention_norm1", "layers.14.adaLN_modulation.0", "layers.10.feed_forward.w2", "layers.9.ffn_norm2", "noise_refiner.1.attention_norm1", "layers.17.ffn_norm1", "layers.1.ffn_norm1", "layers.29.ffn_norm1", "context_refiner.0.feed_forward.w3", "layers.1.ffn_norm2", "layers.1.feed_forward.w2", "layers.1.adaLN_modulation.0", "layers.18.attention_norm1", "layers.24.ffn_norm2", "layers.20.attention.norm_q", "context_refiner.1.attention_norm2", "layers.24.attention_norm2", "layers.22.feed_forward.w1", "layers.23.ffn_norm1", "layers.23.feed_forward.w3", "layers.25.attention.norm_q", "layers.29.attention_norm2", "layers.20.adaLN_modulation.0.bias", "layers.11.feed_forward.w2", "layers.4.feed_forward.w3", "layers.18.feed_forward.w2", "layers.9.attention_norm2", "noise_refiner.1.feed_forward.w2", "layers.17.ffn_norm2", "layers.13.adaLN_modulation.0", "layers.21.adaLN_modulation.0", "layers.25.attention_norm1", "layers.12.ffn_norm2", "layers.13.attention_norm2", "layers.1.attention_norm1", "layers.28.adaLN_modulation.0.bias", "layers.2.adaLN_modulation.0.bias", "layers.12.adaLN_modulation.0.bias", "layers.9.attention.norm_q", "layers.9.feed_forward.w3", "context_refiner.1.attention.norm_q", "layers.19.adaLN_modulation.0", "layers.22.ffn_norm1", "layers.19.ffn_norm1", "layers.10.attention_norm2", "layers.0.attention_norm1", "layers.5.feed_forward.w3", "layers.2.feed_forward.w1", "layers.13.feed_forward.w1", "layers.27.feed_forward.w1", "layers.19.feed_forward.w1", "layers.24.feed_forward.w1", "layers.15.ffn_norm2", "layers.8.adaLN_modulation.0.bias", "layers.14.feed_forward.w2", "layers.14.attention.norm_q", "layers.2.ffn_norm2", "context_refiner.0.ffn_norm1", "layers.7.attention.norm_k", "layers.4.adaLN_modulation.0.bias", "context_refiner.1.feed_forward.w3", "layers.26.attention_norm1", "layers.7.feed_forward.w2", "layers.9.attention.norm_k", "layers.19.adaLN_modulation.0.bias", "layers.10.attention_norm1", "layers.3.adaLN_modulation.0", "layers.4.feed_forward.w2", "layers.8.feed_forward.w1", "layers.8.ffn_norm2", "layers.4.attention_norm2", "layers.5.ffn_norm2", "layers.7.ffn_norm2", "layers.21.attention.norm_k", "layers.19.ffn_norm2", "layers.27.adaLN_modulation.0.bias", "layers.4.ffn_norm1", "layers.15.feed_forward.w2", "all_final_layer.2-1.linear.bias", "layers.22.attention_norm1", "layers.5.adaLN_modulation.0.bias", "layers.2.feed_forward.w3", "layers.21.feed_forward.w3", "layers.11.attention_norm1", "layers.25.feed_forward.w3", "layers.23.attention_norm1", "layers.6.ffn_norm2", "layers.26.attention.norm_q", "layers.18.feed_forward.w3", "context_refiner.0.attention.norm_q", "layers.0.feed_forward.w1", "layers.2.attention.norm_q", "layers.27.feed_forward.w2", "layers.11.adaLN_modulation.0.bias", "layers.26.feed_forward.w3", "layers.8.attention.norm_q", "layers.3.attention_norm2", "layers.16.feed_forward.w3", "layers.7.feed_forward.w3", "layers.8.ffn_norm1", "layers.24.attention.norm_q", "layers.3.attention_norm1", "layers.13.ffn_norm1", "noise_refiner.0.adaLN_modulation.0", "layers.28.attention.norm_q", "layers.29.adaLN_modulation.0", "layers.28.feed_forward.w2", "noise_refiner.1.feed_forward.w3", "context_refiner.0.attention_norm2", "noise_refiner.0.attention.norm_k", "layers.22.adaLN_modulation.0", "layers.23.feed_forward.w1", "context_refiner.1.feed_forward.w2", "layers.9.adaLN_modulation.0.bias", "layers.12.ffn_norm1", "layers.4.adaLN_modulation.0", "t_embedder.mlp.0.bias", "layers.8.attention_norm2", "layers.12.feed_forward.w3", "layers.26.adaLN_modulation.0", "layers.6.adaLN_modulation.0.bias", "layers.26.attention_norm2", "layers.18.attention_norm2", "context_refiner.0.attention.norm_k", "layers.9.feed_forward.w2", "layers.16.feed_forward.w1", "layers.16.adaLN_modulation.0.bias", "layers.13.feed_forward.w2", "layers.1.attention.norm_k", "layers.25.adaLN_modulation.0.bias", "layers.16.attention_norm1", "layers.22.feed_forward.w3", "layers.24.adaLN_modulation.0.bias", "layers.12.attention.norm_q", "layers.6.feed_forward.w1", "layers.14.ffn_norm2", "layers.11.attention_norm2", "layers.3.feed_forward.w1", "noise_refiner.0.feed_forward.w2", "layers.18.attention.norm_k", "layers.18.feed_forward.w1", "layers.23.attention_norm2", "layers.7.adaLN_modulation.0.bias", "t_embedder.mlp.0", "layers.12.feed_forward.w1", "layers.2.attention_norm1", "layers.25.ffn_norm2", "layers.10.ffn_norm2", "layers.5.attention.norm_k", "layers.17.attention.norm_k", "layers.17.attention.norm_q", "layers.6.attention_norm1", "layers.0.feed_forward.w3", "layers.14.attention.norm_k", "layers.19.feed_forward.w3", "layers.10.ffn_norm1", "layers.0.ffn_norm2", "layers.11.ffn_norm1", "noise_refiner.1.attention_norm2", "layers.14.attention_norm2", "layers.3.ffn_norm1", "layers.23.feed_forward.w2", "context_refiner.1.attention.norm_k", "layers.22.attention.norm_q", "layers.17.attention_norm2", "layers.25.attention_norm2", "layers.16.attention.norm_k", "layers.19.attention.norm_k", "layers.2.ffn_norm1", "layers.7.attention.norm_q", "layers.6.feed_forward.w3", "layers.25.adaLN_modulation.0", "layers.13.adaLN_modulation.0.bias", "layers.7.attention_norm2", "layers.15.attention_norm2", "layers.24.attention.norm_k", "layers.2.attention_norm2", "layers.16.attention.norm_q", "layers.20.ffn_norm2", "layers.14.ffn_norm1", "layers.16.adaLN_modulation.0", "layers.19.attention_norm1" ] } }