| { |
| "processor": { |
| "type": "VermoProcessor", |
| "trainable": false, |
| "save_ckpt": false, |
| "module_dtype": "fp32", |
| "pretrained_text_tokenizer": { |
| "type": "AutoTokenizer", |
| "from_pretrained": { |
| "pretrained_model_name_or_path": "tokenizer" |
| } |
| }, |
| "smpl_pose_processor": { |
| "type": "VermoSMPL22Processor", |
| "do_normalize": true, |
| "stats_file": "stats/smplh_universal_stats_aug.json", |
| "rot_type": "rotation_6d", |
| "transl_type": "abs_rel", |
| "smpl_type": "smpl_22" |
| }, |
| "multi_person_smpl_pose_processor": { |
| "type": "VermoSMPL22Processor", |
| "do_normalize": true, |
| "stats_file": "stats/smplh_universal_stats_aug.json", |
| "rot_type": "rotation_6d", |
| "transl_type": "abs_rel", |
| "smpl_type": "smpl_22" |
| }, |
| "motion_tokenizer": { |
| "type": "VQVAEWanMotion2DTK", |
| "from_pretrained": { |
| "pretrained_model_name_or_path": "motion_tokenizer" |
| } |
| }, |
| "audio_tokenizer": null, |
| "audio_codebook_size": 4096, |
| "instruction_stage": true, |
| "optional_input_modal_mode": "none", |
| "task_template_mode": "first", |
| "shuffle_condition_parts": false, |
| "shuffle_output_parts": false, |
| "max_seq_len": 0 |
| }, |
| "lm": { |
| "type": "VermoLlamaForCausalLM", |
| "trainable": false, |
| "save_ckpt": false, |
| "module_dtype": "bf16", |
| "from_pretrained": { |
| "pretrained_model_name_or_path": "lm", |
| "torch_dtype": "bf16", |
| "attn_implementation": "sdpa" |
| } |
| }, |
| "mean_init_embeddings": false |
| } |
|
|
