How to use MoCov3 to conduct pre-training on Custom COCO-style Dataset

[njauwhr]

初次接触mmselfsup
想在自己所做的mmdet的检测任务使用的数据集，使用mocov3做预训练。
目前配置的代码如下

_base_ = [
    '../_base_/models/mocov3_resnet50.py',
    # '../_base_/datasets/imagenet_mocov3.py',
    '../_base_/schedules/lars_coslr-200e_in1k.py',
    '../_base_/default_runtime.py',
]

dataset_type = 'mmdet.CocoDataset'
data_root = '/data1/whrdataset/pidray/'
file_client_args = dict(backend='disk')

train_pipeline = [
    dict(type='LoadImageFromFile', file_client_args=file_client_args),
    dict(
        type='RandomResizedCrop',
        size=224,
        scale=(0.2, 1.0),
        backend='pillow',
        interpolation='bicubic'),
    dict(type='RandomFlip', prob=0.5),
    dict(type='PackSelfSupInputs', meta_keys=['img_path'])
]

train_dataloader = dict(
    batch_size=4,
    num_workers=1,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=True),
    collate_fn=dict(type='default_collate'),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        ann_file='annotations/xray_train.json',
        data_prefix=dict(img='train/'),
        pipeline=train_pipeline))

train_dataloader = dict(batch_size=4, num_workers=1)

# optimizer
optimizer = dict(type='LARS', lr=9.6, weight_decay=1e-6, momentum=0.9)
optim_wrapper = dict(
    type='AmpOptimWrapper',
    loss_scale='dynamic',
    optimizer=optimizer,
    paramwise_cfg=dict(
        custom_keys={
            'bn': dict(decay_mult=0, lars_exclude=True),
            'bias': dict(decay_mult=0, lars_exclude=True),
            # bn layer in ResNet block downsample module
            'downsample.1': dict(decay_mult=0, lars_exclude=True),
        }),
)

# learning rate scheduler
param_scheduler = [
    dict(
        type='LinearLR',
        start_factor=1e-4,
        by_epoch=True,
        begin=0,
        end=10,
        convert_to_iter_based=True),
    dict(
        type='CosineAnnealingLR',
        T_max=90,
        by_epoch=True,
        begin=10,
        end=100,
        convert_to_iter_based=True)
]

# runtime settings
train_cfg = dict(type='EpochBasedTrainLoop', max_epochs=100)

# runtime settings
# only keeps the latest 3 checkpoints
default_hooks = dict(checkpoint=dict(max_keep_ckpts=3))

当使用这个配置文件的时候，在训练的时候，抛出一个异常，认为是数据集的问题，但是一直不知道怎么去改
Result has been saved to /home/ubuntu/whr/mmselfsup-1.x/mmselfsup-1.x/pid/modules_statistic_results.json
11/02 20:38:24 - mmengine - INFO - Distributed training is not used, all SyncBatchNorm (SyncBN) layers in the model will be automatically reverted to BatchNormXd layers if they are used.
Traceback (most recent call last):
File "/home/ubuntu/anaconda3/envs/open-mmlab/lib/python3.8/site-packages/mmengine/registry/build_functions.py", line 121, in build_from_cfg
obj = obj_cls(**args) # type: ignore
File "/home/ubuntu/anaconda3/envs/open-mmlab/lib/python3.8/site-packages/mmengine/runner/loops.py", line 43, in init
super().init(runner, dataloader)
File "/home/ubuntu/anaconda3/envs/open-mmlab/lib/python3.8/site-packages/mmengine/runner/base_loop.py", line 26, in init
self.dataloader = runner.build_dataloader(
File "/home/ubuntu/anaconda3/envs/open-mmlab/lib/python3.8/site-packages/mmengine/runner/runner.py", line 1335, in build_dataloader
dataset_cfg = dataloader_cfg.pop('dataset')
KeyError: 'dataset'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/home/ubuntu/whr/mmselfsup-1.x/mmselfsup-1.x/tools/train.py", line 103, in
main()
File "/home/ubuntu/whr/mmselfsup-1.x/mmselfsup-1.x/tools/train.py", line 99, in main
runner.train()
File "/home/ubuntu/anaconda3/envs/open-mmlab/lib/python3.8/site-packages/mmengine/runner/runner.py", line 1629, in train
self._train_loop = self.build_train_loop(
File "/home/ubuntu/anaconda3/envs/open-mmlab/lib/python3.8/site-packages/mmengine/runner/runner.py", line 1444, in build_train_loop
loop = LOOPS.build(
File "/home/ubuntu/anaconda3/envs/open-mmlab/lib/python3.8/site-packages/mmengine/registry/registry.py", line 421, in build
return self.build_func(cfg, *args, **kwargs, registry=self)
File "/home/ubuntu/anaconda3/envs/open-mmlab/lib/python3.8/site-packages/mmengine/registry/build_functions.py", line 135, in build_from_cfg
raise type(e)(
KeyError: "class EpochBasedTrainLoop in mmengine/runner/loops.py: 'dataset'"

想问问如何解决？
具体环境如图所示

[tonysy]

This question has been resolved by offline communication. Anyone who meets the similar question can try this solution:

• Copy the all content of '../_base_/datasets/imagenet_mocov3.py' into your config and edite the information according to: https://mmselfsup.readthedocs.io/en/dev-1.x/user_guides/4_pretrain_custom_dataset.html