### Install and Setup for Colab

Source: https://github.com/open-mmlab/mmselfsup/blob/main/demo/mae_visualization.ipynb

Installs necessary packages and clones the mmselfsup repository for use in Google Colab. Appends the mmselfsup directory to the system path for import.

```python
import sys

# check whether run in Colab
if 'google.colab' in sys.modules:
    print('Running in Colab.')
    !pip3 install openmim
    !pip install -U openmim
    !mim install mmengine
    !mim install 'mmcv>=2.0.0rc1'

    !git clone https://github.com/open-mmlab/mmselfsup.git
    %cd mmselfsup/
    !git checkout dev-1.x
    !pip install -e .

    sys.path.append('./mmselfsup')
    %cd demo
else:
    sys.path.append('..')
```

--------------------------------

### Install MIM

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/classification.md

Install MIM, a prerequisite for launching benchmarks. Ensure MIM is installed before proceeding.

```shell
pip install openmim
```

--------------------------------

### Example of Browsing a Dataset

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/visualization.md

An example command to run the dataset browsing script with a specific configuration file.

```shell
python tools/misc/browse_dataset.py configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-100e_in1k.py
```

--------------------------------

### Example Training Configuration

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/segmentation.md

An example of a base configuration file for segmentation training, leveraging existing MMSegmentation configurations.

```shell
_base_ = 'mmseg::fcn/fcn_r50-d8_4xb2-40k_cityscapes-769x769.py'
```

--------------------------------

### Example Distributed Training

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/classification.md

An example command for launching distributed training with specific configuration and pre-trained model files.

```shell
bash ./tools/benchmarks/classification/mim_dist_train.sh \
configs/benchmarks/classification/imagenet/resnet50_linear-8xb32-coslr-100e_in1k.py \
work_dir/pretrained_model.pth
```

--------------------------------

### Download and Unzip Example Dataset

Source: https://github.com/open-mmlab/mmselfsup/blob/main/demo/mmselfsup_colab_tutorial.ipynb

Downloads a small example dataset for demonstration purposes. Ensure you have sufficient disk space.

```python
!mkdir data
!wget https://download.openmmlab.com/mmselfsup/data/imagenet_examples.zip
!unzip -q imagenet_examples.zip -d ./data/
```

--------------------------------

### Install Requirements

Source: https://github.com/open-mmlab/mmselfsup/blob/main/projects/fgia_accv2022_1st/README.md

List of software and library versions required for the project. Ensure these versions are installed before proceeding.

```shell
PyTorch 1.11.0
torchvision 0.12.0
CUDA 11.3
MMEngine >= 0.1.0
MMCV >= 2.0.0rc0
MMClassification >= 1.0.0rc1
```

--------------------------------

### Install MMSelfSup from Source

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/get_started.md

Clone the MMSelfSup repository and install it in editable mode. This is recommended for development and customization.

```shell
git clone https://github.com/open-mmlab/mmselfsup.git
cd mmselfsup
pip install -v -e .
```

--------------------------------

### Start Pre-training with Slurm

Source: https://github.com/open-mmlab/mmselfsup/blob/main/projects/fgia_accv2022_1st/README.md

Command to start distributed pre-training using Slurm. This example uses 16 NVIDIA 80G A100 GPUs.

```shell
## we use 16 NVIDIA 80G A100 GPUs for pre-training
GPUS_PER_NODE=8 GPUS=16 SRUN_ARGS=${SRUN_ARGS} bash tools/slurm_train.sh ${PARTITION} ${JOB_NAME} projects/fgia_accv2022_1st/config/mae_vit-large-p16_8xb512-amp-coslr-1600e_in1k.py [optional arguments]
```

--------------------------------

### Example Distributed Testing

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/classification.md

An example command for launching distributed testing with specific configuration and model checkpoint files.

```shell
bash ./tools/benchmarks/mmsegmentation/mim_dist_test.sh \
configs/benchmarks/classification/imagenet/resnet50_linear-8xb32-coslr-100e_in1k.py \
work_dir/model.pth
```

--------------------------------

### Complete Config File Name Example

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/1_config.md

A comprehensive example demonstrating the full config file naming convention by combining all specified parts.

```default
swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96.py
```

--------------------------------

### Install MMEngine and MMCV

Source: https://github.com/open-mmlab/mmselfsup/blob/main/demo/mmselfsup_colab_tutorial.ipynb

Installs the MMEngine and MMCV libraries, which are prerequisites for MMSelfSup. This command ensures you have the correct versions for your environment.

```bash
pip install -v 
    -e git+https://github.com/open-mmlab/mmengine.git#egg=mmengine
    -e git+https://github.com/open-mmlab/mmcv.git#egg=mmcv
```

--------------------------------

### Example Distributed Training Execution

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/segmentation.md

An example command to execute distributed training for segmentation, specifying configuration, pre-trained model, and GPU count.

```shell
bash ./tools/benchmarks/mmsegmentation/mim_dist_train.sh \
configs/benchmarks/mmsegmentation/voc12aug/fcn_r50-d8_4xb4-20k_voc12aug-512x512.py \
https://download.openmmlab.com/mmselfsup/1.x/byol/byol_resnet50_16xb256-coslr-200e_in1k/byol_resnet50_16xb256-coslr-200e_in1k_20220825-de817331.pth 4
```

--------------------------------

### Check MMSelfSup Installation

Source: https://github.com/open-mmlab/mmselfsup/blob/main/demo/mmselfsup_colab_tutorial.ipynb

Verifies the installation of the MMSelfSup library by printing its version. This confirms that the self-supervised learning framework is ready for use.

```python
# Check MMSelfSup installation
import mmselfsup
print(mmselfsup.__version__)
```

--------------------------------

### Install MMEngine and MMCV using mim

Source: https://github.com/open-mmlab/mmselfsup/blob/main/demo/mmselfsup_colab_tutorial.ipynb

Installs the MMEngine and MMCV libraries using the 'mim' package manager. Ensure PyTorch and CUDA versions match the available wheels for optimal installation.

```shell
!pip3 install openmim
!pip install -U openmim
!mim install 'mmengine' 'mmcv>=2.0.0rc4'
```

--------------------------------

### Install MMEngine and MMCV with MIM

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/get_started.md

Use MIM to install the MMEngine and MMCV libraries, which are prerequisites for MMSelfSup.

```shell
pip install -U openmim
mim install mmengine
mim install 'mmcv>=2.0.0'
```

--------------------------------

### Data Information Naming Examples

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/1_config.md

Provides examples of how data information, including dataset name and input image size, is represented in config files.

```default
in1k
```

```default
in1k-384
```

```default
in1k-384x224
```

```default
cifar10
```

```default
inat18
```

```default
places205
```

--------------------------------

### Install and Initialize Pre-commit Hook

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/notes/contribution_guide.md

Install the pre-commit package and initialize the git hook. This ensures that code linters and formatters are enforced on every commit.

```shell
pip install -U pre-commit
```

```shell
pre-commit install
```

```shell
pre-commit run
```

--------------------------------

### Verify MMSelfSup Installation

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/get_started.md

Run this Python code to check if MMSelfSup is installed correctly and print its version.

```python
import mmselfsup
print(mmselfsup.__version__)
```

--------------------------------

### Example Distributed Testing Execution

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/segmentation.md

An example command to execute distributed testing for segmentation, specifying configuration, checkpoint, and GPU count.

```shell
bash ./tools/benchmarks/mmsegmentation/mim_dist_test.sh \
configs/benchmarks/mmsegmentation/voc12aug/fcn_r50-d8_4xb4-20k_voc12aug-512x512.py \
https://download.openmmlab.com/mmselfsup/1.x/byol/byol_resnet50_16xb256-coslr-200e_in1k/byol_resnet50_16xb256-coslr-200e_in1k_20220825-de817331.pth 4
```

--------------------------------

### Start Pre-training with PyTorch (Node 1)

Source: https://github.com/open-mmlab/mmselfsup/blob/main/projects/fgia_accv2022_1st/README.md

Command to start distributed pre-training on the first node using PyTorch. Ensure MASTER_PORT and MASTER_ADDR are set correctly.

```shell
# node 1
NNODES=2 NODE_RANK=0 PORT=${MASTER_PORT} MASTER_ADDR=${MASTER_ADDR} bash tools/dist_train.sh projects/fgia_accv2022_1st/config/mae_vit-large-p16_8xb512-amp-coslr-1600e_in1k.py 8
```

--------------------------------

### Example Distributed Training Command

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/detection.md

A concrete example of initiating distributed training for detection using MMDetection, specifying all necessary arguments.

```shell
bash ./tools/benchmarks/mmdetection/mim_dist_train_c4.sh \
configs/benchmarks/mmdetection/coco/mask-rcnn_r50-c4_ms-1x_coco.py \
https://download.openmmlab.com/mmselfsup/1.x/byol/byol_resnet50_16xb256-coslr-200e_in1k/byol_resnet50_16xb256-coslr-200e_in1k_20220825-de817331.pth 8
```

--------------------------------

### Start Pre-training with PyTorch (Node 2)

Source: https://github.com/open-mmlab/mmselfsup/blob/main/projects/fgia_accv2022_1st/README.md

Command to start distributed pre-training on the second node using PyTorch. Ensure MASTER_PORT and MASTER_ADDR are set correctly.

```shell
# node 2
NNODES=2 NODE_RANK=1 PORT=${MASTER_PORT} MASTER_ADDR=${MASTER_ADDR} bash tools/dist_train.sh projects/fgia_accv2022_1st/config/mae_vit-large-p16_8xb512-amp-coslr-1600e_in1k.py 8
```

--------------------------------

### MAE Pipeline Configuration Example

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/migration.md

Example of a data augmentation pipeline configuration for MAE in MMSelfSup 1.x, demonstrating the structure for image transformations.

```python
train_pipeline = [
    dict(type='LoadImageFromFile'),
    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'])
]

```

--------------------------------

### Example Distributed Testing Command

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/detection.md

A concrete example of initiating distributed testing for a detection model using MMDetection, including configuration and checkpoint paths.

```shell
bash ./tools/benchmarks/mmdetection/mim_dist_test.sh \
configs/benchmarks/mmdetection/coco/mask-rcnn_r50_fpn_ms-1x_coco.py \
https://download.openmmlab.com/mmselfsup/1.x/byol/byol_resnet50_16xb256-coslr-200e_in1k/byol_resnet50_16xb256-coslr-200e_in1k_20220825-de817331.pth 8
```

--------------------------------

### Check MMSelfSup and MMCV Installation

Source: https://github.com/open-mmlab/mmselfsup/blob/main/demo/mmselfsup_colab_tutorial.ipynb

Verifies the installation of mmengine and mmcv by printing their versions and environment details. Ensure PyTorch and CUDA are correctly set up.

```python
# check mmengine install
!python -c 'from mmengine.utils.dl_utils import collect_env;print(collect_env())'

# check mmcv install
import mmcv
print(mmcv.__version__)
```

--------------------------------

### Start Self-Supervised Pre-train Task

Source: https://github.com/open-mmlab/mmselfsup/blob/main/demo/mmselfsup_colab_tutorial.ipynb

This snippet shows the typical logging output during the start of a self-supervised pre-training task. It includes epoch, learning rate, time, data time, memory usage, and loss information.

```log
09/01 16:10:13 - mmengine -  [4m [37mINFO [0m - Epoch(train) [1][80/163]  lr: 2.0000e-01  eta: 0:01:19  time: 0.2033  data_time: 0.0198  memory: 1392  loss: 17.6793
09/01 16:10:15 - mmengine -  [4m [37mINFO [0m - Epoch(train) [1][90/163]  lr: 2.0000e-01  eta: 0:01:12  time: 0.2048  data_time: 0.0195  memory: 1392  loss: 15.4679
09/01 16:10:17 - mmengine -  [4m [37mINFO [0m - Epoch(train) [1][100/163]  lr: 2.0000e-01  eta: 0:01:07  time: 0.2058  data_time: 0.0206  memory: 1392  loss: 6.8410
09/01 16:10:19 - mmengine -  [4m [37mINFO [0m - Epoch(train) [1][110/163]  lr: 2.0000e-01  eta: 0:01:02  time: 0.2036  data_time: 0.0203  memory: 1392  loss: 6.3352
09/01 16:10:21 - mmengine -  [4m [37mINFO [0m - Epoch(train) [1][120/163]  lr: 2.0000e-01  eta: 0:00:58  time: 0.2045  data_time: 0.0200  memory: 1392  loss: 6.0879
09/01 16:10:24 - mmengine -  [4m [37mINFO [0m - Epoch(train) [1][130/163]  lr: 2.0000e-01  eta: 0:00:54  time: 0.2206  data_time: 0.0241  memory: 1392  loss: 4.7499
09/01 16:10:26 - mmengine -  [4m [37mINFO [0m - Epoch(train) [1][140/163]  lr: 2.0000e-01  eta: 0:00:50  time: 0.2143  data_time: 0.0199  memory: 1392  loss: 3.4295
09/01 16:10:28 - mmengine -  [4m [37mINFO [0m - Epoch(train) [1][150/163]  lr: 2.0000e-01  eta: 0:00:47  time: 0.2055  data_time: 0.0199  memory: 1392  loss: 3.2668
09/01 16:10:30 - mmengine -  [4m [37mINFO [0m - Epoch(train) [1][160/163]  lr: 2.0000e-01  eta: 0:00:43  time: 0.2033  data_time: 0.0188  memory: 1392  loss: 2.7335
09/01 16:10:30 - mmengine -  [4m [37mINFO [0m - Exp name: relative-loc_resnet50_8xb64-steplr-70e_in1k_colab_20220901_160940
09/01 16:10:30 - mmengine -  [4m [37mINFO [0m - Saving checkpoint at 1 epochs
09/01 16:10:35 - mmengine -  [4m [37mINFO [0m - Epoch(train) [2][10/163]  lr: 2.0000e-02  eta: 0:00:39  time: 0.2302  data_time: 0.0299  memory: 1392  loss: 2.3706
09/01 16:10:37 - mmengine -  [4m [37mINFO [0m - Epoch(train) [2][20/163]  lr: 2.0000e-02  eta: 0:00:36  time: 0.2050  data_time: 0.0191  memory: 1392  loss: 2.2516
09/01 16:10:39 - mmengine -  [4m [37mINFO [0m - Epoch(train) [2][30/163]  lr: 2.0000e-02  eta: 0:00:33  time: 0.2100  data_time: 0.0202  memory: 1392  loss: 2.2116
09/01 16:10:41 - mmengine -  [4m [37mINFO [0m - Epoch(train) [2][40/163]  lr: 2.0000e-02  eta: 0:00:30  time: 0.2073  data_time: 0.0213  memory: 1392  loss: 2.1653
09/01 16:10:43 - mmengine -  [4m [37mINFO [0m - Epoch(train) [2][50/163]  lr: 2.0000e-02  eta: 0:00:28  time: 0.2103  data_time: 0.0209  memory: 1392  loss: 2.1445
09/01 16:10:45 - mmengine -  [4m [37mINFO [0m - Epoch(train) [2][60/163]  lr: 2.0000e-02  eta: 0:00:25  time: 0.2064  data_time: 0.0190  memory: 1392  loss: 2.1613
09/01 16:10:47 - mmengine -  [4m [37mINFO [0m - Epoch(train) [2][70/163]  lr: 2.0000e-02  eta: 0:00:22  time: 0.2084  data_time: 0.0215  memory: 1392  loss: 2.1216
09/01 16:10:49 - mmengine -  [4m [37mINFO [0m - Epoch(train) [2][80/163]  lr: 2.0000e-02  eta: 0:00:20  time: 0.2060  data_time: 0.0206  memory: 1392  loss: 2.1333
```

--------------------------------

### Install MMCV for PyTorch 1.12.0 and CUDA 11.6

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/get_started.md

Install the mmcv-full package with pip, specifying a find-url based on the PyTorch and CUDA versions. This example installs MMCV built for PyTorch 1.12.0 and CUDA 11.6.

```shell
pip install 'mmcv>=2.0.0rc1' -f https://download.openmmlab.com/mmcv/dist/cu116/torch1.12.0/index.html
```

--------------------------------

### Install MMSegmentation

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/segmentation.md

Install MIM and the MMSegmentation package using pip. Ensure you have MIM installed first.

```shell
pip install openmim
mim install 'mmsegmentation>=1.0.0rc0'
```

--------------------------------

### Install MMEngine with Pip

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/get_started.md

Install MMEngine using pip if you prefer not to use MIM. This is a straightforward installation command.

```shell
pip install mmengine
```

--------------------------------

### Check MMClassification Installation

Source: https://github.com/open-mmlab/mmselfsup/blob/main/demo/mmselfsup_colab_tutorial.ipynb

Verifies the installation of the MMClassification library by printing its version. Ensure MMClassification is installed correctly before proceeding.

```python
# Check MMClassification installation
import mmcls
print(mmcls.__version__)
```

--------------------------------

### Install MMDetection with MIM

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/detection.md

Install MMDetection and its dependencies using the MIM package manager. Ensure MIM is installed first.

```shell
pip install openmim
mim install 'mmdet>=3.0.0rc0'
```

--------------------------------

### Install MMEngine and MMCV in Google Colab

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/get_started.md

Install MMEngine and MMCV using MIM within a Google Colab environment. These commands are necessary before installing MMSelfSup from source.

```shell
!pip3 install openmim
!mim install mmengine
!mim install 'mmcv>=2.0.0rc1'
```

--------------------------------

### Install faiss-gpu with Conda for DeepCluster on A100

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/notes/faq.md

When using DeepCluster on an A100 GPU, installing faiss via pip can cause errors. Use this conda command to install the correct version of faiss-gpu with CUDA 11.3 support. Ensure your PyTorch installation also supports CUDA 11.3, and note that faiss-gpu==1.7.2 requires Python 3.6-3.8.

```bash
conda install -c pytorch faiss-gpu cudatoolkit=11.3
```

--------------------------------

### Install MMSelfSup from Source in Google Colab

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/get_started.md

Clone the MMSelfSup repository, change the directory, checkout the 1.x branch, and install the package in editable mode. This is typically done after installing dependencies in a Google Colab environment.

```shell
!git clone https://github.com/open-mmlab/mmselfsup.git
%cd mmselfsup
!git checkout 1.x
!pip install -e .
```

--------------------------------

### Train with a single GPU

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/3_pretrain.md

Use this command to start training on a single GPU. Replace `${CONFIG_FILE}` with the path to your configuration file.

```shell
python tools/train.py ${CONFIG_FILE} [optional arguments]
```

```shell
python tools/train.py configs/selfsup/mae/mae_vit-base-p16_8xb512-coslr-400e_in1k.py
```

--------------------------------

### Distribute Training with Pre-trained Model

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/4_pretrain_custom_dataset.md

Use this command to initiate distributed training, specifying a pre-trained model checkpoint URL to initialize the model weights. Ensure you replace placeholders like ${CONFIG}, ${GPUS}, and ${PRETRAIN} with your specific values.

```bash
bash tools/dist_train.sh ${CONFIG} ${GPUS} --cfg-options model.pretrained=${PRETRAIN}
```

--------------------------------

### Download and Prepare PASCAL VOC 2007

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/user_guides/2_dataset_prepare.md

This command downloads and prepares the PASCAL VOC 2007 dataset. It assumes your data root is specified by `$YOUR_DATA_ROOT` and creates a symlink for the dataset.

```shell
bash tools/dataset_converters/prepare_voc07_cls.sh $YOUR_DATA_ROOT
```

--------------------------------

### Pre-train on Multiple GPUs

Source: https://github.com/open-mmlab/mmselfsup/blob/main/projects/example_project/README.md

Command to start pre-training on multiple GPUs, specifying the number of GPUs and the launcher. Assumes a PyTorch environment.

```bash
mim train mmselfsup configs/dummy-mae_vit-base-p16_8xb512-amp-coslr-300e_in1k.py \
    --work-dir work_dirs/dummy_mae/ \
    --launcher pytorch --gpus 8
```

--------------------------------

### Initialize and Run MMSelfSup Training

Source: https://github.com/open-mmlab/mmselfsup/blob/main/demo/mmselfsup_colab_tutorial.ipynb

Initializes the training process using MMEngine's Runner by loading a configuration and then starts the training loop. Ensure your configuration object 'cfg' is defined before this.

```python
from mmengine.runner import Runner

# build the runner from config
runner = Runner.from_cfg(cfg)

# start training
runner.train()
```

--------------------------------

### Pre-train on Local Single GPU

Source: https://github.com/open-mmlab/mmselfsup/blob/main/projects/example_project/README.md

Command to initiate pre-training on a single local GPU using a specified configuration file and work directory.

```bash
mim train mmselfsup $CONFIG --work-dir $WORK_DIR
```

```bash
mim train mmselfsup configs/dummy-mae_vit-base-p16_8xb512-amp-coslr-300e_in1k.py \
    --work-dir work_dirs/dummy_mae/
```

--------------------------------

### Install MMSelfSup as a Python Package

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/get_started.md

Install MMSelfSup directly using pip for use in your projects. Ensure you have version '>=1.0.0'.

```shell
pip install 'mmselfsup>=1.0.0'
```

--------------------------------

### Install PyTorch on GPU

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/get_started.md

Install PyTorch and Torchvision for GPU support using conda. Ensure your CUDA toolkit is compatible.

```shell
conda install pytorch torchvision -c pytorch
```

--------------------------------

### BaseModel Initialization in MMSelfSup

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/advanced_guides/models.md

Demonstrates the initialization of `BaseModel`, including optional neck and head modules, and setting up data preprocessors. The `pretrained` argument is used to load pre-trained weights.

```python
class BaseModel(_BaseModel):

    def __init__(self,
                 backbone: dict,
                 neck: Optional[dict] = None,
                 head: Optional[dict] = None,
                 target_generator: Optional[dict] = None,
                 pretrained: Optional[str] = None,
                 data_preprocessor: Optional[Union[dict, nn.Module]] = None,
                 init_cfg: Optional[dict] = None):

        if pretrained is not None:
            init_cfg = dict(type='Pretrained', checkpoint=pretrained)

        if data_preprocessor is None:
            data_preprocessor = {}
        # The build process is in MMEngine, so we need to add scope here.
        data_preprocessor.setdefault('type',
                                     'mmselfsup.SelfSupDataPreprocessor')

        super().__init__(
            init_cfg=init_cfg, data_preprocessor=data_preprocessor)

        self.backbone = MODELS.build(backbone)

        if neck is not None:
            self.neck = MODELS.build(neck)

        if head is not None:
            self.head = MODELS.build(head)

```

--------------------------------

### Configure Visualizer with Multiple Backends

Source: https://github.com/open-mmlab/mmselfsup/blob/main/docs/en/migration.md

Sets up the new `visualizer` component, specifying its type and a list of visualization backends. This allows for flexible logging and visualization of results across different platforms.

```python
visualizer = dict(
    type='SelfSupVisualizer',
    vis_backends=[
        dict(type='LocalVisBackend'),
        # Uncomment the below line to save the log and visualization results to TensorBoard.
        # dict(type='TensorboardVisBackend')
    ]
)
```