{
    "load_video_features": false,
    "load_text_features": false,
    "task": "t2v",
    "ae": {
        "model_id": "hunyuanvideo_15_vae",
        "from_pretrained": "HunyuanVideo1.5/vae/diffusion_pytorch_model.safetensors",
        "dtype": "float16",
        "latent_channels": 32,
        "block_out_channels": [128, 256, 512, 1024, 1024],
        "downsample_match_channel": true,
        "ffactor_spatial": 16,
        "ffactor_temporal": 4,
        "in_channels": 3,
        "layers_per_block": 2,
        "out_channels": 3,
        "sample_size": 128,
        "sample_tsize": 64,
        "scaling_factor": 1.03682,
        "shift_factor": null,
        "upsample_match_channel": true,
        "enable_tiling": true
    },
    "text_encoder": [
        {
            "model_id": "Hunyuan15MLLmModel",
            "dtype": "fp16",
            "from_pretrained": "HunyuanVideo1.5/text_encoder/llm",
            "hub_backend": "hf",
            "use_attention_mask": true,
            "hidden_state_skip_layer": 2,
            "output_key": "hidden_states",
            "template_file_path": "examples/hunyuanvideo_1.5/template.json",
            "template_id": "li-dit-encode-video-json",
            "low_cpu_mem_usage": true
        },
        {
            "model_id": "Hunyuan15GlyphModel",
            "dtype": "fp16",
            "from_pretrained": "HunyuanVideo1.5/text_encoder/byt5-small",
            "hub_backend": "hf",
            "byt5_max_length": 256,
            "hidden_state_skip_layer": 2,
            "output_key": "hidden_states",
            "byT5_ckpt_path": "HunyuanVideo1.5/text_encoder/Glyph-SDXL-v2/checkpoints/byt5_model.pt",
            "color_ann_path": "HunyuanVideo1.5/text_encoder/Glyph-SDXL-v2/assets/color_idx.json",
            "font_ann_path": "HunyuanVideo1.5/text_encoder/Glyph-SDXL-v2/assets/multilingual_10-lang_idx.json"
        }
    ],
    "diffusion": {
        "model_id": "hunyuanvideo_15_diffusion",
        "num_train_timesteps": 1000,
        "shift": 3.0,
        "reverse": true,
        "solver": "euler",
        "snr": "lognorm",
        "train_timestep_shift": 3.0,
        "sp_size": 1,
        "dp_replicate": 1
    },
    "predictor": {
        "model_id": "hunyuanvideo15dit",
        "from_pretrained": null,
        "dtype": "bf16",
        "patch_size": [1, 1, 1],
        "in_channels": 32,
        "concat_condition": true,
        "out_channels": 32,
        "hidden_size": 2048,
        "num_heads": 16,
        "head_dim": 128,
        "mlp_width_ratio": 4,
        "mlp_act_type": "gelu_tanh",
        "mm_double_blocks_depth": 54,
        "mm_single_blocks_depth": 0,
        "rope_dim_list": [16, 56, 56],
        "qkv_bias": true,
        "qk_norm": true,
        "qk_norm_type": "rms",
        "guidance_embed": false,
        "use_meanflow": false,
        "text_projection": "single_refiner",
        "use_attention_mask": true,
        "text_states_dim": 3584,
        "text_states_dim_2": null,
        "text_pool_type": null,
        "rope_theta": 256,
        "attn_mode": "flash2",
        "attn_param": {
            "win_size": [[3, 3, 3]],
            "win_type": "fixed",
            "win_ratio": 10,
            "tile_size": [6, 8, 8],
            "ssta_topk": 64,
            "ssta_threshold": 0.0,
            "ssta_lambda": 0.7,
            "ssta_sampling_type": "importance",
            "ssta_adaptive_pool": null,
            "attn_sparse_type": "ssta",
            "attn_pad_type": "zero",
            "attn_use_text_mask": 1,
            "attn_mask_share_within_head": 0
        },
        "glyph_byT5_v2": true,
        "vision_projection": "linear-none",
        "vision_states_dim": 1152,
        "is_reshape_temporal_channels": false,
        "use_cond_type_embedding": true
    },
    "patch": {
        "clip_grad_async": true,
        "bridge_patch": true
    }
}