{
    "device": "npu",
    "use_ema": false,
    "dtype": "bf16",
    "text_input_ids_pad": true,
    "load_video_features": false,
    "load_text_features": false,
    "allow_internal_format": false,
    "ae": {
        "model_id": "autoencoder_kl_hunyuanvideo",
        "from_pretrained": "Open-Sora-v2/hunyuan_vae.safetensors",
        "dtype": "bf16",
        "latent_channels": 16,
        "block_out_channels": [128, 256, 512, 512],
        "layers_per_block": 2,
        "in_channels": 3,
        "norm_num_groups": 32,
        "out_channels": 3,
        "sample_size": 256,
        "sample_tsize": 64,
        "down_block_types": [
            "DownEncoderBlockCausal3D",
            "DownEncoderBlockCausal3D",
            "DownEncoderBlockCausal3D",
            "DownEncoderBlockCausal3D"
        ],
        "up_block_types": [
            "UpDecoderBlockCausal3D",
            "UpDecoderBlockCausal3D",
            "UpDecoderBlockCausal3D",
            "UpDecoderBlockCausal3D"
        ],
        "scaling_factor": 0.476986,
        "time_compression_ratio": 4,
        "mid_block_add_attention": true,
        "act_fn": "silu",
        "use_spatial_tiling": true,
        "use_temporal_tiling": false
    },
    "text_encoder": [
        {
            "model_id": "T5",
            "hub_backend": "hf",
            "from_pretrained": "Open-Sora-v2/google/t5-v1_1-xxl",
            "dtype": "bf16",
            "use_attention_mask": false,
            "low_cpu_mem_usage": true,
            "ucg_rate": 0.1
        },
        {
            "model_id": "CLIP",
            "hub_backend": "hf",
            "from_pretrained": "Open-Sora-v2/openai/clip-vit-large-patch14",
            "dtype": "bf16",
            "use_attention_mask": false,
            "output_key": "pooler_output",
            "low_cpu_mem_usage": true,
            "ucg_rate": 0.1
        }
    ],
    "predictor": {
        "dtype": "bf16",
        "model_id": "mmdit",
        "from_pretrained": null,
        "patch_size": 2,
        "in_channels": 64,
        "hidden_size": 3072,
        "num_heads": 24,
        "mm_double_blocks_depth": 19,
        "mm_single_blocks_depth": 38,
        "mlp_ratio": 4.0,
        "vec_in_dim": 768,
        "context_in_dim": 4096,
        "axes_dim": [16, 56, 56],
        "attention_q_bias":true,
        "attention_k_bias":true,
        "attention_v_bias":true,
        "fused_qkv": false,
        "guidance_embed": false,
        "double_stream_full_recompute_layers": 19,
        "single_stream_full_recompute_layers": 38,
        "use_liger_rope": true
    },
    "diffusion": {
        "model_id": "opensora2_flow_match_scheduler",
        "use_timestep_transform": true
    }
}