transformer/config.json

{
    "patch_size": [
        1,
        1,
        1
    ],
    "in_channels": 32,
    "concat_condition": true,
    "out_channels": 32,
    "hidden_size": 2048,
    "heads_num": 16,
    "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": "torch_causal",
    "attn_param": null,
    "glyph_byT5_v2": true,
    "vision_projection": "linear",
    "vision_states_dim": 1152,
    "is_reshape_temporal_channels": false,
    "use_cond_type_embedding": true,
    "ideal_resolution": "480p",
    "ideal_task": "i2v",
    "_class_name": "HYWorldTransformer3DModel",
    "_diffusers_version": "0.36.0.dev0"
}