transformer/config.json

{
  "_class_name": "NextDiT",
  "_diffusers_version": "0.30.3",
  "_name_or_path": "/data/input/duongl/finetuning_distributed_multiview_16x8_scalerays_dl3dv_dynamic_shift_softcap_trainingWithFluxScheduler/checkpoint-98000",
  "caption_channels": 2048,
  "decay": 0.9999,
  "depth": 16,
  "ffn_dim_multiplier": null,
  "forward_dtype": "bfloat16",
  "hidden_size": 3840,
  "in_channels": 16,
  "input_size": [
    1,
    96,
    96
  ],
  "inv_gamma": 1.0,
  "min_decay": 0.0,
  "model_max_length": 512,
  "multiple_of": 256,
  "norm_eps": 1e-05,
  "norm_type": "rms",
  "num_heads": 32,
  "num_kv_heads": 8,
  "optimization_step": 8000,
  "param_dtype": "float32",
  "patch_size": [
    1,
    2,
    2
  ],
  "power": 0.6666666666666666,
  "pred_sigma": false,
  "qk_norm": true,
  "remat_attention": "",
  "remat_block": "",
  "remat_mlp": "",
  "rotary_max_length": 128,
  "rotary_max_length_t": null,
  "update_after_step": 0,
  "use_ema_warmup": false
}