(args)
| 1552 | |
| 1553 | |
| 1554 | def build_transformer(args): |
| 1555 | if args.modelname == 'aios_smplx_box': |
| 1556 | return Transformer_Box( |
| 1557 | d_model=args.hidden_dim, |
| 1558 | dropout=args.dropout, |
| 1559 | nhead=args.nheads, |
| 1560 | num_queries=args.num_queries, |
| 1561 | dim_feedforward=args.dim_feedforward, |
| 1562 | num_encoder_layers=args.enc_layers, |
| 1563 | num_decoder_layers=args.dec_layers, |
| 1564 | normalize_before=args.pre_norm, |
| 1565 | return_intermediate_dec=True, |
| 1566 | query_dim=args.query_dim, |
| 1567 | activation=args.transformer_activation, |
| 1568 | num_patterns=args.num_patterns, |
| 1569 | modulate_hw_attn=True, |
| 1570 | deformable_encoder=True, |
| 1571 | deformable_decoder=True, |
| 1572 | num_feature_levels=args.num_feature_levels, |
| 1573 | enc_n_points=args.enc_n_points, |
| 1574 | dec_n_points=args.dec_n_points, |
| 1575 | learnable_tgt_init=True, |
| 1576 | random_refpoints_xy=args.random_refpoints_xy, |
| 1577 | two_stage_type=args.two_stage_type, |
| 1578 | two_stage_learn_wh=args.two_stage_learn_wh, |
| 1579 | two_stage_keep_all_tokens=args.two_stage_keep_all_tokens, |
| 1580 | dec_layer_number=args.dec_layer_number, |
| 1581 | rm_self_attn_layers=args.rm_self_attn_layers, |
| 1582 | rm_detach=args.rm_detach, |
| 1583 | decoder_sa_type=args.decoder_sa_type, |
| 1584 | module_seq=args.decoder_module_seq, |
| 1585 | embed_init_tgt=args.embed_init_tgt, |
| 1586 | num_body_points=args.num_body_points, |
| 1587 | num_hand_points=args.num_hand_points, |
| 1588 | num_face_points=args.num_face_points, |
| 1589 | num_box_decoder_layers=args.num_box_decoder_layers, |
| 1590 | num_hand_face_decoder_layers=args.num_hand_face_decoder_layers, |
| 1591 | num_group=args.num_group) |
| 1592 | elif args.modelname == 'aios_smplx': |
| 1593 | return Transformer( |
| 1594 | d_model=args.hidden_dim, |
| 1595 | dropout=args.dropout, |
| 1596 | nhead=args.nheads, |
| 1597 | num_queries=args.num_queries, |
| 1598 | dim_feedforward=args.dim_feedforward, |
| 1599 | num_encoder_layers=args.enc_layers, |
| 1600 | num_decoder_layers=args.dec_layers, |
| 1601 | normalize_before=args.pre_norm, |
| 1602 | return_intermediate_dec=True, |
| 1603 | query_dim=args.query_dim, |
| 1604 | activation=args.transformer_activation, |
| 1605 | num_patterns=args.num_patterns, |
| 1606 | modulate_hw_attn=True, |
| 1607 | deformable_encoder=True, |
| 1608 | deformable_decoder=True, |
| 1609 | num_feature_levels=args.num_feature_levels, |
| 1610 | enc_n_points=args.enc_n_points, |
| 1611 | dec_n_points=args.dec_n_points, |
no test coverage detected