configs/bbox_gen.yaml

model:
  name: bbox_gen
  args:
    encoder_dim_feat: 448
    encoder_dim: 64
    encoder_heads: 4
    encoder_token_num: 2048
    encoder_qkv_bias: false
    encoder_use_ln_post: true
    encoder_use_checkpoint: true
    encoder_num_embed_freqs: 8
    encoder_embed_include_pi: false
    encoder_init_scale: 0.25
    encoder_random_fps: true
    encoder_learnable_query: true
    encoder_layers: 8

    max_group_size: 50

    vocab_size: 67
    decoder_hidden_size: 1024
    decoder_num_hidden_layers: 24
    decoder_ffn_dim: 4096
    decoder_heads: 16
    decoder_use_flash_attention: true
    decoder_gradient_checkpointing: false

    bins: 64
    BOS_id: 64
    EOS_id: 65
    PAD_id: 66
    max_length: 2187
    voxel_token_length: 1886
    voxel_token_placeholder: -1