Text to Animation
Released
Type a prompt, get an animation. Text-to-motion for any skeleton (humans, animals, creatures)
Editor/TextToAnimation/Inference/UniMate/UniMateParamOrder.cs
Static generated file that lists the parameter names and shapes for the UniMate model version v2, exposed as a readonly array of (Name, Shape) tuples. It is used to map checkpoint shadow parameters to expected tensor shapes and ordering.
// Generated from UniMate (Linzhan/UniMate @ 92710b30) unimate_uniml3d_f60_v2: model.named_parameters() order,
// which is the order of ema_state_dict["shadow_params"] in the checkpoint. Do not edit by hand.
namespace TextToAnimation.Editor.Inference.UniMate;
internal static class UniMateParamOrder
{
/// <summary>(name, shape) per shadow parameter index.</summary>
public static readonly (string Name, int[] Shape)[] V2 =
{
("cond_embedder.weight", new[] { 512,768 }),
("cond_embedder.bias", new[] { 512 }),
("time_embedder.mlp.0.weight", new[] { 512,256 }),
("time_embedder.mlp.0.bias", new[] { 512 }),
("time_embedder.mlp.2.weight", new[] { 512,512 }),
("time_embedder.mlp.2.bias", new[] { 512 }),
("input_layer.root_tpos_embedder.0.weight", new[] { 512,12 }),
("input_layer.root_tpos_embedder.0.bias", new[] { 512 }),
("input_layer.root_tpos_embedder.2.weight", new[] { 512,512 }),
("input_layer.root_tpos_embedder.2.bias", new[] { 512 }),
("input_layer.root_x_embedder.0.weight", new[] { 512,12 }),
("input_layer.root_x_embedder.0.bias", new[] { 512 }),
("input_layer.root_x_embedder.2.weight", new[] { 512,512 }),
("input_layer.root_x_embedder.2.bias", new[] { 512 }),
("input_layer.joint_tpos_embedder.0.weight", new[] { 512,12 }),
("input_layer.joint_tpos_embedder.0.bias", new[] { 512 }),
("input_layer.joint_tpos_embedder.2.weight", new[] { 512,512 }),
("input_layer.joint_tpos_embedder.2.bias", new[] { 512 }),
("input_layer.parent_tpos_embedder.0.weight", new[] { 512,12 }),
("input_layer.parent_tpos_embedder.0.bias", new[] { 512 }),
("input_layer.parent_tpos_embedder.2.weight", new[] { 512,512 }),
("input_layer.parent_tpos_embedder.2.bias", new[] { 512 }),
("input_layer.tpos_fuse.0.weight", new[] { 512,1024 }),
("input_layer.tpos_fuse.0.bias", new[] { 512 }),
("input_layer.tpos_fuse.2.weight", new[] { 512,512 }),
("input_layer.tpos_fuse.2.bias", new[] { 512 }),
("input_layer.joint_x_embedder.0.weight", new[] { 512,12 }),
("input_layer.joint_x_embedder.0.bias", new[] { 512 }),
("input_layer.joint_x_embedder.2.weight", new[] { 512,512 }),
("input_layer.joint_x_embedder.2.bias", new[] { 512 }),
("tpos_pool.pool.queries", new[] { 1,4,512 }),
("tpos_pool.pool.norm_q.weight", new[] { 512 }),
("tpos_pool.pool.norm_kv.weight", new[] { 512 }),
("tpos_pool.pool.cross_attn.in_proj_weight", new[] { 1536,512 }),
("tpos_pool.pool.cross_attn.in_proj_bias", new[] { 1536 }),
("tpos_pool.pool.cross_attn.out_proj.weight", new[] { 512,512 }),
("tpos_pool.pool.cross_attn.out_proj.bias", new[] { 512 }),
("tpos_pool.pool.norm_ff.weight", new[] { 512 }),
("tpos_pool.pool.ffn.w12.weight", new[] { 2048,512 }),
("tpos_pool.pool.ffn.w12.bias", new[] { 2048 }),
("tpos_pool.pool.ffn.w3.weight", new[] { 512,1024 }),
("tpos_pool.pool.ffn.w3.bias", new[] { 512 }),
("tpos_pool.pool.norm_out.weight", new[] { 512 }),
("tpos_pool.pool.out_proj.weight", new[] { 512,512 }),
("tpos_pool.pool.out_proj.bias", new[] { 512 }),
("joint_name_embedder.weight", new[] { 512,768 }),
("joint_name_embedder.bias", new[] { 512 }),
("depth_embedding.weight", new[] { 20,512 }),
("rope_j.spectral_encoder.phi.0.weight", new[] { 64,1 }),
("rope_j.spectral_encoder.phi.0.bias", new[] { 64 }),
("rope_j.spectral_encoder.phi.2.weight", new[] { 64,64 }),
("rope_j.spectral_encoder.phi.2.bias", new[] { 64 }),
("rope_j.spectral_encoder.rho.0.weight", new[] { 64,512 }),
("rope_j.spectral_encoder.rho.0.bias", new[] { 64 }),
("rope_j.spectral_encoder.rho.2.weight", new[] { 32,64 }),
("rope_j.spectral_encoder.rho.2.bias", new[] { 32 }),
("transformer_blocks.0.norm_s.weight", new[] { 512 }),
("transformer_blocks.0.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.0.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.0.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.0.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.0.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.0.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.0.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.0.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.0.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.0.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.0.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.0.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.0.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.0.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.0.norm_t.weight", new[] { 512 }),
("transformer_blocks.0.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.0.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.0.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.0.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.0.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.0.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.0.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.0.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.0.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.0.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.0.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.0.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.0.adaLN_modulation.1.bias", new[] { 4608 }),
("transformer_blocks.1.norm_s.weight", new[] { 512 }),
("transformer_blocks.1.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.1.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.1.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.1.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.1.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.1.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.1.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.1.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.1.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.1.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.1.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.1.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.1.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.1.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.1.norm_t.weight", new[] { 512 }),
("transformer_blocks.1.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.1.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.1.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.1.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.1.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.1.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.1.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.1.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.1.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.1.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.1.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.1.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.1.adaLN_modulation.1.bias", new[] { 4608 }),
("transformer_blocks.2.norm_s.weight", new[] { 512 }),
("transformer_blocks.2.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.2.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.2.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.2.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.2.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.2.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.2.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.2.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.2.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.2.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.2.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.2.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.2.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.2.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.2.norm_t.weight", new[] { 512 }),
("transformer_blocks.2.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.2.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.2.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.2.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.2.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.2.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.2.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.2.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.2.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.2.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.2.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.2.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.2.adaLN_modulation.1.bias", new[] { 4608 }),
("transformer_blocks.3.norm_s.weight", new[] { 512 }),
("transformer_blocks.3.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.3.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.3.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.3.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.3.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.3.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.3.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.3.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.3.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.3.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.3.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.3.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.3.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.3.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.3.norm_t.weight", new[] { 512 }),
("transformer_blocks.3.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.3.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.3.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.3.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.3.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.3.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.3.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.3.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.3.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.3.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.3.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.3.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.3.adaLN_modulation.1.bias", new[] { 4608 }),
("transformer_blocks.4.norm_s.weight", new[] { 512 }),
("transformer_blocks.4.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.4.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.4.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.4.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.4.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.4.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.4.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.4.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.4.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.4.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.4.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.4.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.4.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.4.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.4.norm_t.weight", new[] { 512 }),
("transformer_blocks.4.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.4.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.4.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.4.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.4.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.4.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.4.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.4.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.4.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.4.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.4.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.4.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.4.adaLN_modulation.1.bias", new[] { 4608 }),
("transformer_blocks.5.norm_s.weight", new[] { 512 }),
("transformer_blocks.5.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.5.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.5.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.5.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.5.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.5.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.5.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.5.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.5.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.5.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.5.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.5.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.5.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.5.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.5.norm_t.weight", new[] { 512 }),
("transformer_blocks.5.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.5.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.5.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.5.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.5.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.5.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.5.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.5.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.5.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.5.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.5.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.5.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.5.adaLN_modulation.1.bias", new[] { 4608 }),
("transformer_blocks.6.norm_s.weight", new[] { 512 }),
("transformer_blocks.6.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.6.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.6.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.6.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.6.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.6.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.6.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.6.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.6.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.6.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.6.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.6.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.6.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.6.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.6.norm_t.weight", new[] { 512 }),
("transformer_blocks.6.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.6.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.6.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.6.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.6.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.6.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.6.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.6.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.6.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.6.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.6.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.6.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.6.adaLN_modulation.1.bias", new[] { 4608 }),
("transformer_blocks.7.norm_s.weight", new[] { 512 }),
("transformer_blocks.7.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.7.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.7.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.7.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.7.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.7.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.7.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.7.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.7.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.7.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.7.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.7.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.7.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.7.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.7.norm_t.weight", new[] { 512 }),
("transformer_blocks.7.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.7.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.7.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.7.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.7.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.7.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.7.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.7.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.7.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.7.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.7.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.7.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.7.adaLN_modulation.1.bias", new[] { 4608 }),
("transformer_blocks.8.norm_s.weight", new[] { 512 }),
("transformer_blocks.8.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.8.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.8.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.8.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.8.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.8.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.8.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.8.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.8.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.8.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.8.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.8.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.8.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.8.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.8.norm_t.weight", new[] { 512 }),
("transformer_blocks.8.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.8.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.8.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.8.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.8.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.8.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.8.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.8.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.8.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.8.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.8.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.8.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.8.adaLN_modulation.1.bias", new[] { 4608 }),
("transformer_blocks.9.norm_s.weight", new[] { 512 }),
("transformer_blocks.9.s_attn.graph_dist_scale", new[] { 1 }),
("transformer_blocks.9.s_attn.graph_rel_scale", new[] { 1 }),
("transformer_blocks.9.s_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.9.s_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.9.s_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.9.s_attn.proj.bias", new[] { 512 }),
("transformer_blocks.9.s_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.9.s_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.9.s_attn.graph_dist_embedding.weight", new[] { 6,128 }),
("transformer_blocks.9.s_attn.graph_dist_proj.weight", new[] { 8,128 }),
("transformer_blocks.9.s_attn.graph_dist_proj.bias", new[] { 8 }),
("transformer_blocks.9.s_attn.graph_rel_embedding.weight", new[] { 6,128 }),
("transformer_blocks.9.s_attn.graph_rel_proj.weight", new[] { 8,128 }),
("transformer_blocks.9.s_attn.graph_rel_proj.bias", new[] { 8 }),
("transformer_blocks.9.norm_t.weight", new[] { 512 }),
("transformer_blocks.9.t_attn.qkv.weight", new[] { 1536,512 }),
("transformer_blocks.9.t_attn.qkv.bias", new[] { 1536 }),
("transformer_blocks.9.t_attn.proj.weight", new[] { 512,512 }),
("transformer_blocks.9.t_attn.proj.bias", new[] { 512 }),
("transformer_blocks.9.t_attn.q_norm.weight", new[] { 64 }),
("transformer_blocks.9.t_attn.k_norm.weight", new[] { 64 }),
("transformer_blocks.9.norm_mlp.weight", new[] { 512 }),
("transformer_blocks.9.mlp.w12.weight", new[] { 2730,512 }),
("transformer_blocks.9.mlp.w12.bias", new[] { 2730 }),
("transformer_blocks.9.mlp.w3.weight", new[] { 512,1365 }),
("transformer_blocks.9.mlp.w3.bias", new[] { 512 }),
("transformer_blocks.9.adaLN_modulation.1.weight", new[] { 4608,512 }),
("transformer_blocks.9.adaLN_modulation.1.bias", new[] { 4608 }),
("final_layer.norm_final.weight", new[] { 512 }),
("final_layer.root_cross_attn.in_proj_weight", new[] { 1536,512 }),
("final_layer.root_cross_attn.in_proj_bias", new[] { 1536 }),
("final_layer.root_cross_attn.out_proj.weight", new[] { 512,512 }),
("final_layer.root_cross_attn.out_proj.bias", new[] { 512 }),
("final_layer.root_cross_norm_q.weight", new[] { 512 }),
("final_layer.root_cross_norm_kv.weight", new[] { 512 }),
("final_layer.root_out.0.weight", new[] { 512,512 }),
("final_layer.root_out.0.bias", new[] { 512 }),
("final_layer.root_out.2.weight", new[] { 12,512 }),
("final_layer.root_out.2.bias", new[] { 12 }),
("final_layer.joint_out.0.weight", new[] { 512,512 }),
("final_layer.joint_out.0.bias", new[] { 512 }),
("final_layer.joint_out.2.weight", new[] { 12,512 }),
("final_layer.joint_out.2.bias", new[] { 12 }),
("final_layer.adaLN_modulation.1.weight", new[] { 1024,512 }),
("final_layer.adaLN_modulation.1.bias", new[] { 1024 }),
};
}