About data_loader in configs

[1nyourlife]

I am trying to fine-tuning models in mmdetection on my dataset, based on official document, I try to define data_loader like this:

val_dataloader = dict(
batch_size=1,
num_workers=2,
dataset=dict(
type=dataset_type,
test_mode=True,
# explicitly add your class names to the field metainfo
metainfo=dict(classes=classes),
data_root=data_root,
ann_file='annotations/annotation_val_coco.json',
data_prefix=dict(img='val2017/')
)
)

And successfully reproduce the faster-rcnn and train it on my dataset.

However, instead of define the data_loader directly, I found that many of configs define a specific train_pipeline to load the dataset, and I can not find where they define and register the pipeline.

My question is where can I find the definition of the built-in pipelines, and where can I register a custom pipeline. Can I reproduce other models by directly define a dataloader like above?

Please feel free to discuss with me in Chinese.