Config System¶

We use python files as our config system. You can find all the provided configs under $MMEditing/configs.

Config Name Style¶

We follow the below style to name config files. Contributors are advised to follow the same style.

{model}_[model setting]_{backbone}_[refiner]_[norm setting]_[misc]_[gpu x batch_per_gpu]_{schedule}_{dataset}

{xxx} is required field and [yyy] is optional.

{model}: model type like srcnn, dim, etc.
[model setting]: specific setting for some model, like resolution for input images, stage name for training, etc.
{backbone}: backbone type like r50 (ResNet-50), x101 (ResNeXt-101).
[refiner]: refiner type like pln (Plain Refiner).
[norm_setting]: bn (Batch Normalization) is used unless specified, other norm layer type could be gn (Group Normalization), syncbn (Synchronized Batch Normalization).
[misc]: miscellaneous setting/plugins of model, e.g. dconv, gcb, attention, albu, mstrain.
[gpu x batch_per_gpu]: GPUs and samples per GPU, 8x2 is used by default.
{schedule}: training schedule, 20k, 100k, etc. 20k means 20,000 iterations. 100k means 100,000 iterations.
{dataset}: dataset like places (for inpainting), comp1k (for matting), div2k (for restoration) and paired (for generation).

Config File Structure¶

Please refer to the corresponding page for config file structure for different tasks.

FAQ¶

Use intermediate variables in configs¶

Some intermediate variables are used in the configs files, like train_pipeline/test_pipeline in datasets.

For example, we would like to first define the train_pipeline/test_pipeline and pass them into data. Thus, train_pipeline/test_pipeline are intermediate variable.

...
train_dataset_type = 'SRAnnotationDataset'
val_dataset_type = 'SRFolderDataset'
train_pipeline = [
    dict(
        type='LoadImageFromFile',
        io_backend='disk',
        key='lq',
        flag='unchanged'),
    ...
    dict(type='Collect', keys=['lq', 'gt'], meta_keys=['lq_path', 'gt_path']),
    dict(type='ImageToTensor', keys=['lq', 'gt'])
]
test_pipeline = [
    dict(
        type='LoadImageFromFile',
        io_backend='disk',
        key='lq',
        flag='unchanged'),
    ...
    dict(type='Collect', keys=['lq', 'gt'], meta_keys=['lq_path', 'lq_path']),
    dict(type='ImageToTensor', keys=['lq', 'gt'])
]

data = dict(
    # train
    samples_per_gpu=16,
    workers_per_gpu=6,
    drop_last=True,
    train=dict(
        type='RepeatDataset',
        times=1000,
        dataset=dict(
            type=train_dataset_type,
            lq_folder='data/DIV2K/DIV2K_train_LR_bicubic/X2_sub',
            gt_folder='data/DIV2K/DIV2K_train_HR_sub',
            ann_file='data/DIV2K/meta_info_DIV2K800sub_GT.txt',
            pipeline=train_pipeline,
            scale=scale)),
    # val
    val_samples_per_gpu=1,
    val_workers_per_gpu=1,
    val=dict(
        type=val_dataset_type,
        lq_folder='./data/val_set5/Set5_bicLRx2',
        gt_folder='./data/val_set5/Set5_mod12',
        pipeline=test_pipeline,
        scale=scale,
        filename_tmpl='{}'),