{
    "load_video_features": false,
    "load_text_features": false,
    "task": "i2v",
    "diffusion": {
        "model_id": "wan_flow_match_scheduler",
        "num_train_timesteps": 1000,
        "shift": 5,
        "sigma_min": 0.0,
        "extra_one_step": true,
        "max_timestep_boundary": 0.358,
        "min_timestep_boundary": 0.0
    },
    "predictor": {
        "model_id": "wandit",
        "dtype": "bf16",
        "model_type": "wan2.2-i2v",
        "patch_size": [1, 2, 2],
        "text_len": 512,
        "in_dim": 36,
        "hidden_size": 5120,
        "ffn_dim": 13824,
        "freq_dim": 256,
        "text_dim": 4096,
        "img_dim": 1280,
        "out_dim": 16,
        "num_heads": 40,
        "num_layers": 40,
        "pipeline_num_layers": [10,10,10,10],
        "qk_norm": true,
        "qk_norm_type": "rmsnorm",
        "cross_attn_norm": true,
        "eps": 1e-6,
        "max_seq_len": 1024,
        "attention_async_offload": false,
        "use_fused_rmsnorm":true
    },
    "patch": {
        "clip_grad_async": true,
        "get_dist_model_load_from_pt": true
    },
    "encoder_offload_interval": 8,
    "ae": {
        "model_id": "wan_video_vae",
        "from_pretrained": "Wan2.2-I2V-A14B-Diffusers/vae",
        "dtype": "bf16",
        "enable_tiling": false,
        "tiling_param": {
            "tile_sample_min_height": 256,
            "tile_sample_min_width": 256,
            "tile_sample_stride_height": 192,
            "tile_sample_stride_width": 192
        },
        "norm_latents": true,
        "norm_mode": "channel_specified_shift_scale",
        "do_sample": false,
        "i2v_processor": {
            "processor_id": "wan_i2v_processor",
            "i2v_vae_encod_tiling": false
        }
    },
    "text_encoder": {
        "model_id": "UMT5",
        "dtype": "bf16",
        "hub_backend": "hf",
        "from_pretrained": "Wan2.2-I2V-A14B-Diffusers/text_encoder"
    }
}