default_config.yaml

#1. Configuring the Running Information
device_id: 0
device_target: 'GPU'
context_mode: 'GRAPH_MODE' #'PYNATIVE_MODE'
#2. Processing Data
dataset_path: "/datasets/ImageNet/train/"
repeat_num: 1
batch_size: 32
run_distribute: False
num_parallel_workers: 6
#3. Defining the Network

#4. Defining the Loss Function and Optimizer
lr_init: 0.0
lr_end: 0.0001
lr_max: 0.1
warmup_epochs: 0
lr_decay_mode: "step"

momentum: 0.9
loss_scale: 1024

label_smooth_factor: 0.1
class_num: 1000


#5. Training the Network
save_ckpt_path: "./ckpt/"
save_ckpt_step: 1000
keep_checkpoint: 3
num_epoch: 1
eval_during_training: True
eval_path: "/datasets/ImageNet/val/"
eval_per_epoch: 1
summary_recorder_path: "./summary_recorder/"

#6. evaluation.py->test_net
checkpoint_name: './ckpt/checkpoint_darknet53-1_20018.ckpt'



# # Builtin Configurations(DO NOT CHANGE THESE CONFIGURATIONS unless you know exactly what you are doing)
# enable_modelarts: False
# # Url for modelarts
# data_url: ""
# train_url: ""
# checkpoint_url: ""
# # Path for local
# data_path: "/cache/data"
# output_path: "/cache/train"
# load_path: "/cache/checkpoint_path"
# device_target: "Ascend" # ['Ascend', 'GPU']
# need_modelarts_dataset_unzip: True
# modelarts_dataset_unzip_name: "coco2014"

# # ==============================================================================
# # Training options

# # dataset related
# data_dir: "/cache/data/coco2014/"
# per_batch_size: 32

# # network related
# pretrained_backbone: "/cache/checkpoint_path/0-148_92000.ckpt"
# resume_yolov3: ""

# # optimizer and lr related
# lr_scheduler: "exponential"
# lr: 0.001
# lr_epochs: "220,250"
# lr_gamma: 0.1
# eta_min: 0.0
# T_max: 320
# max_epoch: 320
# warmup_epochs: 0
# weight_decay: 0.0005
# momentum: 0.9

# # loss related
# loss_scale: 1024
# label_smooth: 0
# label_smooth_factor: 0.1

# # logging related
# log_interval: 1
# ckpt_path: "outputs/"
# ckpt_interval: -1
# is_save_on_master: 1

# # distributed related
# is_distributed: 1
# rank: 0
# group_size: 1

# # profiler init
# need_profiler: 0

# # reset default config
# training_shape: ""


# # Eval option
# pretrained: ""
# log_path: "outputs/"
# nms_thresh: 0.5
# annFile: ""
# testing_shape: ""
# eval_ignore_threshold: 0.001


# # Export option
# device_id: 0
# batch_size: 1
# ckpt_file: ""
# file_name: "yolov3_darknet53"
# file_format: "AIR" # ["AIR", "ONNX", "MINDIR"]
# keep_detect: True

# # PostProcess option
# result_path: ""
# img_path: ""

# # convert weight option
# input_file: "./darknet53.conv.74"
# output_file: "./backbone_darknet53.ckpt"

# # Other default config
# hue: 0.1
# saturation: 1.5
# value: 1.5
# jitter: 0.3

# resize_rate: 1
# multi_scale: [[320, 320],
#               [352, 352],
#               [384, 384],
#               [416, 416],
#               [448, 448],
#               [480, 480],
#               [512, 512],
#               [544, 544],
#               [576, 576],
#               [608, 608]
#               ]

# num_classes: 80
# out_channel: 255 #3 * (num_classes + 5)
# max_box: 50

# backbone_input_shape: [32, 64, 128, 256, 512]
# backbone_shape: [64, 128, 256, 512, 1024]
# backbone_layers: [1, 2, 8, 8, 4]

# # confidence under ignore_threshold means no object when training
# ignore_threshold: 0.7

# # h->w
# anchor_scales: [[10, 13],
#                 [16, 30],
#                 [33, 23],
#                 [30, 61],
#                 [62, 45],
#                 [59, 119],
#                 [116, 90],
#                 [156, 198],
#                 [373, 326]]

# # test_param
# test_img_shape: [416, 416]

# ---

# # Help description for each configuration
# data_dir: "Train dataset directory."
# per_batch_size: "Batch size for Training."
# pretrained_backbone: "The ckpt file of DarkNet53."
# resume_yolov3: "The ckpt file of YOLOv3, which used to fine tune."

# lr_scheduler: "Learning rate scheduler, options: exponential, cosine_annealing."
# lr: "Learning rate."
# lr_epochs: "Epoch of changing of lr changing, split with ',' ."
# lr_gamma: "Decrease lr by a factor of exponential lr_scheduler."
# eta_min: "Eta_min in cosine_annealing scheduler."
# T_max: "T-max in cosine_annealing scheduler."
# max_epoch: "Max epoch num to train the model."
# warmup_epochs: "Warmup epochs."
# weight_decay: "Weight decay factor."
# momentum: "Momentum."
# loss_scale: "Static loss scale."
# label_smooth: "Whether to use label smooth in CE."
# label_smooth_factor: "Smooth strength of original one-hot."
# log_interval: "Logging interval steps."
# ckpt_path: "Checkpoint save location."
# ckpt_interval: "Save checkpoint interval."
# is_save_on_master: "Save ckpt on master or all rank, 1 for master, 0 for all ranks."
# is_distributed: "Distribute train or not, 1 for yes, 0 for no."
# rank: "Local rank of distributed."
# group_size: "World size of device."
# need_profiler: "Whether use profiler. 0 for no, 1 for yes."
# training_shape: "Fix training shape."
# resize_rate: "Resize rate for multi-scale training."

# # eval option
# pretrained: "model_path, local pretrained model to load."
# log_path: "checkpoint save location."
# nms_thresh: "threshold for NMS."
# annFile: "path to annotation."
# testing_shape: "shape for test."
# eval_ignore_threshold: "threshold to throw low quality boxes for eval."

# # export option
# device_id: "Device id"
# batch_size: "batch size"
# ckpt_file: "Checkpoint file path."
# file_name: "output file name."
# file_format: "file format choices in ['AIR', 'ONNX', 'MINDIR']"
# device_target: "device target. choices in ['Ascend', 'GPU'] for train. choices in ['Ascend', 'GPU', 'CPU'] for export."
# keep_detect: "keep the detect module or not, default: True"

# # convert weight option
# input_file: "input file path."
# output_file: "output file path."