| { | |
| "model_type": "mdm", | |
| "format": "motius-mdm-v2", | |
| "guidance_param": 2.5, | |
| "components": { | |
| "clip": { | |
| "name": "ViT-B/32", | |
| "stored_in_artifact": true, | |
| "path": "clip.safetensors" | |
| } | |
| }, | |
| "config": { | |
| "dataset": "humanml", | |
| "arch": "trans_enc", | |
| "latent_dim": 512, | |
| "layers": 8, | |
| "cond_mode": "text", | |
| "cond_mask_prob": 0.1, | |
| "emb_trans_dec": false, | |
| "noise_schedule": "cosine", | |
| "diffusion_steps": 1000, | |
| "sigma_small": true, | |
| "text_encoder_type": "clip", | |
| "pos_embed_max_len": 5000, | |
| "mask_frames": false, | |
| "emb_policy": "add", | |
| "pred_len": 0, | |
| "context_len": 0, | |
| "multi_target_cond": false, | |
| "multi_encoder_type": "multi", | |
| "target_enc_layers": 1, | |
| "lambda_vel": 0.0, | |
| "lambda_rcxyz": 0.0, | |
| "lambda_fc": 0.0, | |
| "unconstrained": false | |
| } | |
| } |