yacs的使用教程

原文鏈接：https://blog.csdn.net/u013066730/article/details/97640131

例一

新建一個experiment.yaml的文件

# my_project/experiment.yaml
SYSTEM:
  NUM_GPUS: 2
TRAIN:
  SCALES: (1, 2)

然後新建一個python文件

from config import get_cfg_defaults
# local variable usage pattern, or:
# from config import cfg  # global singleton usage pattern
 
 
if __name__ == "__main__":
  cfg = get_cfg_defaults()
  cfg.merge_from_file("experiment.yaml")
  print("GPUS:"+str(cfg.SYSTEM.NUM_GPUS))
  cfg.freeze()
  print(cfg)

結果爲 

GPUS:2
SYSTEM:
  NUM_GPUS: 2
  NUM_WORKERS: 4
TRAIN:
  HYPERPARAMETER_1: 0.1
  SCALES: (1, 2)

​

 其中get_cfg_defaults，相當於構建了一個容器，可以將子一級變成父一級的變量（也就是變成父類中的一個變量），只需要使用 . 這個符號就能訪問該變量。

例二

新建一個hrnet.yaml文件、

CUDNN:
  BENCHMARK: true
  DETERMINISTIC: false
  ENABLED: true
GPUS: (0,1,2,3)
OUTPUT_DIR: 'output'
LOG_DIR: 'log'
WORKERS: 4
PRINT_FREQ: 100
 
DATASET:
  DATASET: cityscapes
  ROOT: 'data/'
  TEST_SET: 'list/cityscapes/val.lst'
  TRAIN_SET: 'list/cityscapes/train.lst'
  NUM_CLASSES: 19
MODEL:
  NAME: seg_hrnet
  PRETRAINED: 'pretrained_models/hrnetv2_w48_imagenet_pretrained.pth'
  EXTRA:
    FINAL_CONV_KERNEL: 1
    STAGE2:
      NUM_MODULES: 1
      NUM_BRANCHES: 2
      BLOCK: BASIC
      NUM_BLOCKS:
      - 4
      - 4
      NUM_CHANNELS:
      - 48
      - 96
      FUSE_METHOD: SUM
    STAGE3:
      NUM_MODULES: 4
      NUM_BRANCHES: 3
      BLOCK: BASIC
      NUM_BLOCKS:
      - 4
      - 4
      - 4
      NUM_CHANNELS:
      - 48
      - 96
      - 192
      FUSE_METHOD: SUM
    STAGE4:
      NUM_MODULES: 3
      NUM_BRANCHES: 4
      BLOCK: BASIC
      NUM_BLOCKS:
      - 4
      - 4
      - 4
      - 4
      NUM_CHANNELS:
      - 48
      - 96
      - 192
      - 384
      FUSE_METHOD: SUM
LOSS:
  USE_OHEM: false
  OHEMTHRES: 0.9
  OHEMKEEP: 131072
TRAIN:
  IMAGE_SIZE:
  - 1024
  - 512
  BASE_SIZE: 2048
  BATCH_SIZE_PER_GPU: 3
  SHUFFLE: true
  BEGIN_EPOCH: 0
  END_EPOCH: 484
  RESUME: true
  OPTIMIZER: sgd
  LR: 0.01
  WD: 0.0005
  MOMENTUM: 0.9
  NESTEROV: false
  FLIP: true
  MULTI_SCALE: true
  DOWNSAMPLERATE: 1
  IGNORE_LABEL: 255
  SCALE_FACTOR: 16
TEST:
  IMAGE_SIZE:
  - 2048
  - 1024
  BASE_SIZE: 2048
  BATCH_SIZE_PER_GPU: 4
  FLIP_TEST: false
  MULTI_SCALE: false

然後新建一個python文件，

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
 
import os
 
from yacs.config import CfgNode as CN
 
_C = CN()
 
_C.OUTPUT_DIR = ''
_C.LOG_DIR = ''
_C.GPUS = (0,)
_C.WORKERS = 4
_C.PRINT_FREQ = 20
_C.AUTO_RESUME = False
_C.PIN_MEMORY = True
_C.RANK = 0
 
# Cudnn related params
_C.CUDNN = CN()
_C.CUDNN.BENCHMARK = True
_C.CUDNN.DETERMINISTIC = False
_C.CUDNN.ENABLED = True
 
# common params for NETWORK
_C.MODEL = CN()
_C.MODEL.NAME = 'seg_hrnet'
_C.MODEL.PRETRAINED = ''
_C.MODEL.EXTRA = CN(new_allowed=True)
 
_C.LOSS = CN()
_C.LOSS.USE_OHEM = False
_C.LOSS.OHEMTHRES = 0.9
_C.LOSS.OHEMKEEP = 100000
_C.LOSS.CLASS_BALANCE = False
 
# DATASET related params
_C.DATASET = CN()
_C.DATASET.ROOT = ''
_C.DATASET.DATASET = 'cityscapes'
_C.DATASET.NUM_CLASSES = 19
_C.DATASET.TRAIN_SET = 'list/cityscapes/train.lst'
_C.DATASET.EXTRA_TRAIN_SET = ''
_C.DATASET.TEST_SET = 'list/cityscapes/val.lst'
 
# training
_C.TRAIN = CN()
 
_C.TRAIN.IMAGE_SIZE = [1024, 512]  # width * height
_C.TRAIN.BASE_SIZE = 2048
_C.TRAIN.DOWNSAMPLERATE = 1
_C.TRAIN.FLIP = True
_C.TRAIN.MULTI_SCALE = True
_C.TRAIN.SCALE_FACTOR = 16
 
_C.TRAIN.LR_FACTOR = 0.1
_C.TRAIN.LR_STEP = [90, 110]
_C.TRAIN.LR = 0.01
_C.TRAIN.EXTRA_LR = 0.001
 
_C.TRAIN.OPTIMIZER = 'sgd'
_C.TRAIN.MOMENTUM = 0.9
_C.TRAIN.WD = 0.0001
_C.TRAIN.NESTEROV = False
_C.TRAIN.IGNORE_LABEL = -1
 
_C.TRAIN.BEGIN_EPOCH = 0
_C.TRAIN.END_EPOCH = 484
_C.TRAIN.EXTRA_EPOCH = 0
 
_C.TRAIN.RESUME = False
 
_C.TRAIN.BATCH_SIZE_PER_GPU = 32
_C.TRAIN.SHUFFLE = True
# only using some training samples
_C.TRAIN.NUM_SAMPLES = 0
 
# testing
_C.TEST = CN()
 
_C.TEST.IMAGE_SIZE = [2048, 1024]  # width * height
_C.TEST.BASE_SIZE = 2048
 
_C.TEST.BATCH_SIZE_PER_GPU = 32
# only testing some samples
_C.TEST.NUM_SAMPLES = 0
 
_C.TEST.MODEL_FILE = ''
_C.TEST.FLIP_TEST = False
_C.TEST.MULTI_SCALE = False
_C.TEST.SCALE_LIST = [1]
 
# debug
_C.DEBUG = CN()
_C.DEBUG.DEBUG = False
_C.DEBUG.SAVE_BATCH_IMAGES_GT = False
_C.DEBUG.SAVE_BATCH_IMAGES_PRED = False
_C.DEBUG.SAVE_HEATMAPS_GT = False
_C.DEBUG.SAVE_HEATMAPS_PRED = False
 
 
 
if __name__ == '__main__':
    cfg = _C
 
    cfg.defrost()
 
    cfg.merge_from_file("hrnet.yaml")
    print(cfg.TEST.IMAGE_SIZE)
    cfg.freeze()
    print(cfg)
 

結果爲：

[2048, 1024]
AUTO_RESUME: False
CUDNN:
  BENCHMARK: True
  DETERMINISTIC: False
  ENABLED: True
DATASET:
  DATASET: cityscapes
  EXTRA_TRAIN_SET: 
  NUM_CLASSES: 19
  ROOT: data/
  TEST_SET: list/cityscapes/val.lst
  TRAIN_SET: list/cityscapes/train.lst
DEBUG:
  DEBUG: False
  SAVE_BATCH_IMAGES_GT: False
  SAVE_BATCH_IMAGES_PRED: False
  SAVE_HEATMAPS_GT: False
  SAVE_HEATMAPS_PRED: False
GPUS: (0, 1, 2, 3)
LOG_DIR: log
LOSS:
  CLASS_BALANCE: False
  OHEMKEEP: 131072
  OHEMTHRES: 0.9
  USE_OHEM: False
MODEL:
  EXTRA:
    FINAL_CONV_KERNEL: 1
    STAGE2:
      BLOCK: BASIC
      FUSE_METHOD: SUM
      NUM_BLOCKS: [4, 4]
      NUM_BRANCHES: 2
      NUM_CHANNELS: [48, 96]
      NUM_MODULES: 1
    STAGE3:
      BLOCK: BASIC
      FUSE_METHOD: SUM
      NUM_BLOCKS: [4, 4, 4]
      NUM_BRANCHES: 3
      NUM_CHANNELS: [48, 96, 192]
      NUM_MODULES: 4
    STAGE4:
      BLOCK: BASIC
      FUSE_METHOD: SUM
      NUM_BLOCKS: [4, 4, 4, 4]
      NUM_BRANCHES: 4
      NUM_CHANNELS: [48, 96, 192, 384]
      NUM_MODULES: 3
  NAME: seg_hrnet
  PRETRAINED: pretrained_models/hrnetv2_w48_imagenet_pretrained.pth
OUTPUT_DIR: output
PIN_MEMORY: True
PRINT_FREQ: 100
RANK: 0
TEST:
  BASE_SIZE: 2048
  BATCH_SIZE_PER_GPU: 4
  FLIP_TEST: False
  IMAGE_SIZE: [2048, 1024]
  MODEL_FILE: 
  MULTI_SCALE: False
  NUM_SAMPLES: 0
  SCALE_LIST: [1]
TRAIN:
  BASE_SIZE: 2048
  BATCH_SIZE_PER_GPU: 3
  BEGIN_EPOCH: 0
  DOWNSAMPLERATE: 1
  END_EPOCH: 484
  EXTRA_EPOCH: 0
  EXTRA_LR: 0.001
  FLIP: True
  IGNORE_LABEL: 255
  IMAGE_SIZE: [1024, 512]
  LR: 0.01
  LR_FACTOR: 0.1
  LR_STEP: [90, 110]
  MOMENTUM: 0.9
  MULTI_SCALE: True
  NESTEROV: False
  NUM_SAMPLES: 0
  OPTIMIZER: sgd
  RESUME: True
  SCALE_FACTOR: 16
  SHUFFLE: True
  WD: 0.0005
WORKERS: 4

這裏要注意一個重要的點：在_C中申明過後的變量才能被賦值和修改，如果_C中沒有某個變量，而yaml中有，那就會報錯，因爲該變量未能事先申明。

但從_C.MODEL.EXTRA = CN(new_allowed=True)這句代碼可以看出，我們申明瞭父類的變量，而且允許子類新建，那麼yaml中如下部分可以被申明。

MODEL:
  EXTRA:
    FINAL_CONV_KERNEL: 1
    STAGE2:
      BLOCK: BASIC
      FUSE_METHOD: SUM
      NUM_BLOCKS: [4, 4]
      NUM_BRANCHES: 2
      NUM_CHANNELS: [48, 96]
      NUM_MODULES: 1
    STAGE3:
      BLOCK: BASIC
      FUSE_METHOD: SUM
      NUM_BLOCKS: [4, 4, 4]
      NUM_BRANCHES: 3
      NUM_CHANNELS: [48, 96, 192]
      NUM_MODULES: 4
    STAGE4:
      BLOCK: BASIC
      FUSE_METHOD: SUM
      NUM_BLOCKS: [4, 4, 4, 4]
      NUM_BRANCHES: 4
      NUM_CHANNELS: [48, 96, 192, 384]
      NUM_MODULES: 3