| { | |
| "architectures": ["PixDiT_T2I"], | |
| "model_type": "pixeldit", | |
| "in_channels": 3, | |
| "patch_size": 16, | |
| "num_groups": 24, | |
| "hidden_size": 1536, | |
| "pixel_hidden_size": 16, | |
| "pixel_attn_hidden_size": 1152, | |
| "pixel_num_groups": 16, | |
| "patch_depth": 14, | |
| "pixel_depth": 2, | |
| "num_text_blocks": 4, | |
| "txt_embed_dim": 2304, | |
| "txt_max_length": 300, | |
| "use_text_rope": true, | |
| "text_rope_theta": 10000.0, | |
| "repa_encoder_index": 6, | |
| "use_pixel_abs_pos": true, | |
| "image_size": 1024, | |
| "text_encoder": "gemma-2-2b-it", | |
| "num_parameters": 1300000000, | |
| "scheduler": { | |
| "type": "flow_matching", | |
| "flow_shift": 4.0, | |
| "weighting_scheme": "logit_normal", | |
| "logit_mean": 0.0, | |
| "logit_std": 1.0 | |
| }, | |
| "sampling": { | |
| "algorithm": "flow_dpm-solver", | |
| "default_steps": 50, | |
| "default_cfg_scale": 2.75, | |
| "default_negative_prompt": "low quality, worst quality, over-saturated, blurry, deformed, watermark" | |
| } | |
| } | |