configs/generation/ss_flow_txt_dit_XL_16l8_fp16.json

{
    "models": {
        "denoiser": {
            "name": "SparseStructureFlowModel",
            "args": {
                "resolution": 16,
                "in_channels": 8,
                "out_channels": 8,
                "model_channels": 1280,
                "cond_channels": 768,
                "num_blocks": 28,
                "num_heads": 16,
                "mlp_ratio": 4,
                "patch_size": 1,
                "pe_mode": "ape",
                "qk_rms_norm": true,
                "qk_rms_norm_cross": true,
                "use_fp16": true
            }
        }
    },
    "dataset": {
        "name": "TextConditionedSparseStructureLatent",
        "args": {
            "latent_model": "ss_enc_conv3d_16l8_fp16",
            "min_aesthetic_score": 4.5,
            "pretrained_ss_dec": "microsoft/TRELLIS-image-large/ckpts/ss_dec_conv3d_16l8_fp16"
        }
    },
    "trainer": {
        "name": "TextConditionedFlowMatchingCFGTrainer",
        "args": {
            "max_steps": 1000000,
            "batch_size_per_gpu": 4,
            "batch_split": 1,
            "optimizer": {
                "name": "AdamW",
                "args": {
                    "lr": 0.0001,
                    "weight_decay": 0.0
                }
            },
            "ema_rate": [
                0.9999
            ],
            "fp16_mode": "inflat_all",
            "fp16_scale_growth": 0.001,
            "grad_clip": {
                "name": "AdaptiveGradClipper",
                "args": {
                    "max_norm": 1.0,
                    "clip_percentile": 95
                }
            },
            "i_log": 500,
            "i_sample": 10000,
            "i_save": 10000,
            "p_uncond": 0.1,
            "t_schedule": {
                "name": "logitNormal",
                "args": {
                    "mean": 1.0,
                    "std": 1.0
                }
            },
            "sigma_min": 1e-5,
            "text_cond_model": "openai/clip-vit-large-patch14"
        }
    }
}