default.yml 2.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121
  1. data:
  2. batch_size: 256
  3. pin_memory: true
  4. num_workers: 6
  5. # Thomas said it should be at least about 5-10x your batch size; beyond that,
  6. # the differences become academic.
  7. shuffle_buffer: 10000
  8. seed: ${train.seed}
  9. dataset:
  10. meta:
  11. gcc3m:
  12. type: img_txt_pair
  13. path: local_data/gcc3m_shards
  14. prefix: gcc-train-{000000..00436}.tar
  15. length: 2891445
  16. gcc12m:
  17. type: img_txt_pair
  18. path: local_data/gcc12m_shards
  19. prefix: gcc-conceptual-12m-{000000..001943}.tar
  20. length: 11156203
  21. yfcc14m:
  22. type: img_txt_pair
  23. path: local_data/yfcc14m_shards
  24. prefix: yfcc14m-{000000..001888}.tar
  25. length: 14615499
  26. redcap12m:
  27. type: img_txt_pair
  28. path: local_data/redcap12m_shards
  29. prefix: redcap12m-{000000..001211}.tar
  30. length: 11866987
  31. imagenet:
  32. type: img_cls_pair
  33. path: local_data/imagenet_shards
  34. prefix: imagenet-val-{000000..000049}.tar
  35. length: 50000
  36. cuhkpedes_train:
  37. type: img_txt_pair
  38. path: local_data/cuhkpedes_shards
  39. prefix: cuhkpedes-train-{000000..000255}.tar
  40. length: 34054
  41. cuhkpedes_val:
  42. type: img_txt_pair
  43. path: local_data/cuhkpedes_shards
  44. prefix: cuhkpedes-val-{000000..000023}.tar
  45. length: 3078
  46. train:
  47. # - gcc3m
  48. - gcc12m
  49. - yfcc14m
  50. - cuhkpedes_train
  51. val:
  52. # - imagenet
  53. - cuhkpedes_val
  54. img_aug:
  55. deit_aug: true
  56. img_size: 224
  57. img_scale: [0.08, 1.0]
  58. interpolation: bilinear
  59. color_jitter: 0.4
  60. auto_augment: 'rand-m9-mstd0.5-inc1'
  61. re_prob: 0.25
  62. re_mode: 'pixel'
  63. re_count: 1
  64. text_aug:
  65. max_seq_len: 77
  66. multi_label: 0
  67. word_type: 'noun'
  68. train:
  69. start_epoch: 0
  70. epochs: 30
  71. warmup_epochs: 2
  72. base_lr: 1.6e-3
  73. weight_decay: 0.05
  74. warmup_lr: 4e-6
  75. min_lr: 4e-5
  76. clip_grad: 5.0
  77. accumulation_steps: 0
  78. amp_opt_level: O2
  79. seed: 0
  80. lr_scheduler:
  81. name: cosine
  82. optimizer:
  83. name: adamw
  84. eps: 1e-8
  85. betas: [0.9, 0.999]
  86. evaluate:
  87. eval_only: false
  88. eval_freq: 1
  89. task:
  90. - cls
  91. - seg
  92. cls:
  93. save_best: true
  94. template: subset
  95. seg:
  96. save_best: true
  97. cfg: segmentation/configs/_base_/datasets/pascal_voc12.py
  98. template: simple
  99. opts: []
  100. checkpoint:
  101. auto_resume: true
  102. resume: ''
  103. freq: 1
  104. max_kept: -1
  105. save_freq: 1
  106. model_name: '' # display name in the logger
  107. output: ???
  108. tag: default
  109. print_freq: 10
  110. seed: 0
  111. wandb: false
  112. local_rank: ???
  113. vis: []