ldm_b/decoder/config.json

{
  "_class_name": "LVAEDecoderWrapper",
  "_diffusers_version": "0.29.2",
  "act_fn": "silu",
  "activation": "relu",
  "apply_gt_mask": true,
  "bce_weight_matching": 6.0,
  "bce_weight_objective_high": 5.0,
  "bce_weight_objective_low": 1.0,
  "bce_weight_objective_medium": 3.0,
  "block_out_channels": [
    64,
    128,
    256,
    512
  ],
  "construction_big_color": [
    70,
    70,
    70
  ],
  "construction_small_color": [
    102,
    102,
    156
  ],
  "d_ffn": 1024,
  "d_model": 512,
  "dec_layers": 6,
  "dim_feedforward": 1024,
  "dist_fcn": "center_distance",
  "dist_th_tp": 2.0,
  "dist_ths": [
    0.5,
    1.0,
    2.0,
    4.0
  ],
  "double_z": false,
  "down_block_types": [
    "DownEncoderBlock2D",
    "DownEncoderBlock2D",
    "DownEncoderBlock2D",
    "DownEncoderBlock2D"
  ],
  "dropout": 0.0,
  "dropout_prob": 0.0,
  "enc_layers": 0,
  "encoder_dim": 512,
  "grid_shape": [
    256,
    256,
    32
  ],
  "ground_color": [
    81,
    0,
    81
  ],
  "head_d_ffn": 1024,
  "head_num_layers": 3,
  "human_color": [
    220,
    20,
    60
  ],
  "in_channels": 10,
  "l1_center_weight_matching": 3.0,
  "l1_center_weight_objective_high": 15.0,
  "l1_center_weight_objective_low": 3.0,
  "l1_center_weight_objective_medium": 9.0,
  "l1_cholesky_weight_matching": 3.0,
  "l1_cholesky_weight_objective_high": 12.0,
  "l1_cholesky_weight_objective_low": 2.0,
  "l1_cholesky_weight_objective_medium": 7.0,
  "lambda_height": 9.0,
  "lambda_kl": 1e-06,
  "lambda_mask": 1.0,
  "latent_channels": 64,
  "latent_frame": [
    32,
    32
  ],
  "layers_per_block": 1,
  "max_boxes_per_sample": 500,
  "mean_ap_weight": 5,
  "mid_block_add_attention": true,
  "min_precision": 0.1,
  "min_recall": 0.1,
  "n_heads": 8,
  "norm_by_count": false,
  "norm_by_count_new": true,
  "norm_num_groups": 32,
  "normalize_before": false,
  "normalize_by_all": false,
  "num_common_blocks": 4,
  "num_construction_big_queries": 16,
  "num_construction_small_queries": 77,
  "num_encoder_layers": 6,
  "num_heads": 8,
  "num_human_queries": 5,
  "num_object_queries": 17,
  "num_pole_queries": 19,
  "num_traffic_control_queries": 17,
  "num_two_wheelers_queries": 6,
  "num_vegetation_cuboid_queries": 178,
  "num_vegetation_ellipsoid_queries": 159,
  "num_vehicle_big_queries": 2,
  "num_vehicle_small_queries": 18,
  "object_color": [
    0,
    128,
    192
  ],
  "out_channels": [
    5,
    5
  ],
  "parking_color": [
    250,
    170,
    160
  ],
  "patch_size": 1,
  "pixel_frame": [
    256,
    256
  ],
  "plot_examples": 0,
  "pole_color": [
    153,
    153,
    153
  ],
  "position_embedding": "sine",
  "reconstruction_loss": "l1",
  "render_curves": false,
  "road_color": [
    128,
    64,
    128
  ],
  "sample_size": 32,
  "sidewalk_color": [
    244,
    35,
    232
  ],
  "split_latent": true,
  "terrain_color": [
    152,
    251,
    152
  ],
  "threshold": 0.3,
  "traffic_control_color": [
    250,
    170,
    30
  ],
  "two_wheelers_color": [
    119,
    11,
    32
  ],
  "up_block_types": [
    "UpDecoderBlock2D",
    "UpDecoderBlock2D",
    "UpDecoderBlock2D",
    "UpDecoderBlock2D"
  ],
  "use_post_quant_conv": true,
  "use_quant_conv": true,
  "vegetation_cuboid_color": [
    107,
    142,
    35
  ],
  "vegetation_ellipsoid_color": [
    107,
    142,
    35
  ],
  "vehicle_big_color": [
    0,
    60,
    100
  ],
  "vehicle_small_color": [
    0,
    0,
    142
  ],
  "verbose": false,
  "voxel_size": 0.25,
  "voxel_z_offset": 0.5
}