TransLO/configs.py at main · IRMVLab/TransLO · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
# -*- coding:UTF-8 -*-

import argparse


"""
  config
"""

def translonet_args():

    parser = argparse.ArgumentParser()
    parser.add_argument('--gpu', type=int, default=0, help='GPU to use [default: GPU 3]')
    parser.add_argument('--multi_gpu', type=str, default=None, help='The gpu [default : null]')
    parser.add_argument('--limit_or_filter', type=bool, default=True, help='if False, filter will reserve 40m~50m points')
    parser.add_argument('--batch_size', type=int, default=8, help='Batch Size during training [default: 16]')
    parser.add_argument('--eval_batch_size', type=int, default=8, help='Batch Size during evaling [default: 64]')
    parser.add_argument('--eval_before', type=int, default=0, help='if 1, eval before train')

    parser.add_argument('--lidar_root', default='/dataset/data_odometry_velodyne/dataset', help='Dataset directory [default: /dataset]')
    parser.add_argument('--image_root', default='/dataset/data_odometry_color', help='Dataset directory [default: /dataset]')
    parser.add_argument('--log_dir', default='log_train', help='Log dir [default: log_train]')

    parser.add_argument('--num_points', type=int, default=150000, help='Point Number [default: 2048]')

    parser.add_argument('--H_input', type=int, default=64, help='H Number [default: 64]')
    parser.add_argument('--W_input', type=int, default=1792, help='W Number [default: 1800]')

    parser.add_argument('--max_epoch', type=int, default=301, help='Epoch to run [default: 151]')
    parser.add_argument('--weight_decay', type=int, default=0.0001, help='The Weight decay [default : 0.0001]')
    parser.add_argument('--workers', type=int, default=6,
                        help='Sets how many child processes can be used [default : 16]')
    parser.add_argument('--model_name', type=str, default='pwclonet', help='base_dir_name [default: pwclonet]')
    parser.add_argument('--task_name', type=str, default=None, help='who can replace model_name ')
    parser.add_argument('--learning_rate', type=float, default=0.001, help='Initial learning rate [default: 0.001]')

    parser.add_argument('--ckpt', type=str, default=None)
    parser.add_argument('--optimizer', default='Adam', help='adam or momentum [default: adam]')

    parser.add_argument('--initial_lr', type=bool, default=False, help='Initial learning rate or not [default: False]')
    parser.add_argument('--learning_rate_clip', type=float, default=1e-6, help='learning_rate_clip [default : 1e-5]')
    parser.add_argument('--lr_stepsize', type=int, default=13, help="lr_stepsize")
    parser.add_argument('--lr_gamma', type=float, default=0.7, help="lr_gamma")
    parser.add_argument('--momentum', type=float, default=0.9, help='Initial learning rate [default: 0.9]')
    parser.add_argument('--is_training', type=bool, default=True, help='is_training [default : True]')

    ##Trnasformers
    #Encoder
    parser.add_argument('--d_embed', type=int, default=128, help='Number of dimensions to encode into')
    parser.add_argument('--scale', type=float, default=1.0, help='for pos embedding')
    parser.add_argument('--attention_type', type=str, default='dot_prod', help='attention type')
    parser.add_argument('--nhead', type=int, default=8, help='heads of Multi-Head Attention')
    parser.add_argument('--d_feedforward', type=int, default=512, help='')
    parser.add_argument('--dropout', type=float, default=0.0, help='drop out')
    parser.add_argument('--H_anchors', type=int, default=4, help='size of H patches')
    parser.add_argument('--W_anchors', type=int, default=4, help='size of W patches')
    parser.add_argument('--seq_len', type=int, default=1, help='sequence length')
    parser.add_argument('--pre_norm', type=bool, default=True, help='Normalization type')
    parser.add_argument('--transformer_act', type=str, default='relu', help='transformer activation type')
    parser.add_argument('--num_encoder_layers', type=int, default=6, help='number of encoder layers')
    parser.add_argument('--transformer_encoder_has_pos_emb', type=bool, default=True,
                        help='if transformer encoder has pos emb')
    parser.add_argument('--sa_val_has_pos_emb', type=bool, default=True, help='if f1 has pos emb')
    parser.add_argument('--ca_val_has_pos_emb', type=bool, default=True, help='if f2 has pos emb')
    #Decoder
    parser.add_argument('--corr_decoder_has_pos_emb', type=bool, default=True, help='if decoder has pos emb')


    args = parser.parse_args()
    return args