base_options.py 7.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105
  1. import argparse
  2. import os
  3. from util import util
  4. import torch
  5. class BaseOptions():
  6. def __init__(self):
  7. self.parser = argparse.ArgumentParser()
  8. self.initialized = False
  9. def initialize(self):
  10. # experiment specifics
  11. self.parser.add_argument('--name', type=str, default='people', help='name of the experiment. It decides where to store samples and models')
  12. self.parser.add_argument('--gpu_ids', type=str, default='0', help='gpu ids: e.g. 0 0,1,2, 0,2. use -1 for CPU')
  13. self.parser.add_argument('--checkpoints_dir', type=str, default='./checkpoints', help='models are saved here')
  14. self.parser.add_argument('--model', type=str, default='pix2pixHD', help='which model to use')
  15. self.parser.add_argument('--norm', type=str, default='batch', help='instance normalization or batch normalization')
  16. self.parser.add_argument('--use_dropout', action='store_true', help='use dropout for the generator')
  17. self.parser.add_argument('--data_type', default=32, type=int, choices=[8, 16, 32], help="Supported data type i.e. 8, 16, 32 bit")
  18. self.parser.add_argument('--verbose', action='store_true', default=False, help='toggles verbose')
  19. self.parser.add_argument('--fp16', action='store_true', default=False, help='train with AMP')
  20. self.parser.add_argument('--local_rank', type=int, default=0, help='local rank for distributed training')
  21. self.parser.add_argument('--isTrain', type=bool, default=True, help='local rank for distributed training')
  22. # input/output sizes
  23. self.parser.add_argument('--batchSize', type=int, default=8, help='input batch size')
  24. self.parser.add_argument('--loadSize', type=int, default=1024, help='scale images to this size')
  25. self.parser.add_argument('--fineSize', type=int, default=512, help='then crop to this size')
  26. self.parser.add_argument('--label_nc', type=int, default=0, help='# of input label channels')
  27. self.parser.add_argument('--input_nc', type=int, default=3, help='# of input image channels')
  28. self.parser.add_argument('--output_nc', type=int, default=3, help='# of output image channels')
  29. # for setting inputs
  30. self.parser.add_argument('--dataroot', type=str, default='./datasets/cityscapes/')
  31. self.parser.add_argument('--resize_or_crop', type=str, default='scale_width', help='scaling and cropping of images at load time [resize_and_crop|crop|scale_width|scale_width_and_crop]')
  32. self.parser.add_argument('--serial_batches', action='store_true', help='if true, takes images in order to make batches, otherwise takes them randomly')
  33. self.parser.add_argument('--no_flip', action='store_true', help='if specified, do not flip the images for data argumentation')
  34. self.parser.add_argument('--nThreads', default=2, type=int, help='# threads for loading data')
  35. self.parser.add_argument('--max_dataset_size', type=int, default=float("inf"), help='Maximum number of samples allowed per dataset. If the dataset directory contains more than max_dataset_size, only a subset is loaded.')
  36. # for displays
  37. self.parser.add_argument('--display_winsize', type=int, default=512, help='display window size')
  38. self.parser.add_argument('--tf_log', action='store_true', help='if specified, use tensorboard logging. Requires tensorflow installed')
  39. # for generator
  40. self.parser.add_argument('--netG', type=str, default='global', help='selects model to use for netG')
  41. self.parser.add_argument('--latent_size', type=int, default=512, help='latent size of Adain layer')
  42. self.parser.add_argument('--ngf', type=int, default=64, help='# of gen filters in first conv layer')
  43. self.parser.add_argument('--n_downsample_global', type=int, default=3, help='number of downsampling layers in netG')
  44. self.parser.add_argument('--n_blocks_global', type=int, default=6, help='number of residual blocks in the global generator network')
  45. self.parser.add_argument('--n_blocks_local', type=int, default=3, help='number of residual blocks in the local enhancer network')
  46. self.parser.add_argument('--n_local_enhancers', type=int, default=1, help='number of local enhancers to use')
  47. self.parser.add_argument('--niter_fix_global', type=int, default=0, help='number of epochs that we only train the outmost local enhancer')
  48. # for instance-wise features
  49. self.parser.add_argument('--no_instance', action='store_true', help='if specified, do *not* add instance map as input')
  50. self.parser.add_argument('--instance_feat', action='store_true', help='if specified, add encoded instance features as input')
  51. self.parser.add_argument('--label_feat', action='store_true', help='if specified, add encoded label features as input')
  52. self.parser.add_argument('--feat_num', type=int, default=3, help='vector length for encoded features')
  53. self.parser.add_argument('--load_features', action='store_true', help='if specified, load precomputed feature maps')
  54. self.parser.add_argument('--n_downsample_E', type=int, default=4, help='# of downsampling layers in encoder')
  55. self.parser.add_argument('--nef', type=int, default=16, help='# of encoder filters in the first conv layer')
  56. self.parser.add_argument('--n_clusters', type=int, default=10, help='number of clusters for features')
  57. self.parser.add_argument('--image_size', type=int, default=224, help='number of clusters for features')
  58. self.parser.add_argument('--norm_G', type=str, default='spectralspadesyncbatch3x3', help='instance normalization or batch normalization')
  59. self.parser.add_argument('--semantic_nc', type=int, default=3, help='number of clusters for features')
  60. self.initialized = True
  61. def parse(self, save=True):
  62. if not self.initialized:
  63. self.initialize()
  64. self.opt = self.parser.parse_args()
  65. self.opt.isTrain = self.isTrain # train or test
  66. str_ids = self.opt.gpu_ids.split(',')
  67. self.opt.gpu_ids = []
  68. for str_id in str_ids:
  69. id = int(str_id)
  70. print("gpuID:"+str_id)
  71. if id >= 0:
  72. self.opt.gpu_ids.append(id)
  73. # set gpu ids
  74. if len(self.opt.gpu_ids) > 0:
  75. torch.cuda.set_device(self.opt.gpu_ids[0])
  76. args = vars(self.opt)
  77. print('------------ Options -------------')
  78. for k, v in sorted(args.items()):
  79. print('%s: %s' % (str(k), str(v)))
  80. print('-------------- End ----------------')
  81. # save to the disk
  82. if self.opt.isTrain:
  83. expr_dir = os.path.join(self.opt.checkpoints_dir, self.opt.name)
  84. util.mkdirs(expr_dir)
  85. if save and not self.opt.continue_train:
  86. file_name = os.path.join(expr_dir, 'opt.txt')
  87. with open(file_name, 'wt') as opt_file:
  88. opt_file.write('------------ Options -------------\n')
  89. for k, v in sorted(args.items()):
  90. opt_file.write('%s: %s\n' % (str(k), str(v)))
  91. opt_file.write('-------------- End ----------------\n')
  92. return self.opt