| { | |
| "camera_attn_class": { | |
| "_partial_": true, | |
| "_target_": "vggttt.nets.vggt.layers.attention.Attention" | |
| }, | |
| "global_attn_class": { | |
| "_partial_": true, | |
| "_target_": "vggttt.nets.ttt_attention.FastWeightAttention", | |
| "base_lr": 0.01, | |
| "div_lr_by_seq_len": false, | |
| "mlp_ratio": 4, | |
| "muon_update_steps": 5, | |
| "num_steps": 2, | |
| "short_conv_size_qkv": [ | |
| 0, | |
| 0, | |
| 3 | |
| ] | |
| }, | |
| "gradient_checkpoint": true, | |
| "img_size": 518, | |
| "init_weights": "VGGT-1B", | |
| "use_point_head": true, | |
| "use_track_head": false | |
| } |