[Improve] Speed up data preprocessor.

[mzr1996]

Motivation

The original data preprocessor consumes much time because it needs to cast the label tensors of samples one by one.

Modification

1. In ClsDataPreprocessoor, only cast gt_label of data samples to cuda instead of the whole data sample. And before casting, stack/concat the tensors.
2. Refactoring the batch augmentations class like Mixup. Now it directly processes the batch inputs and batch scores (one-hot format labels), and won't accept normal labels or data samples.
3. Add ClsDataSample serialization override functions. During serialization in ForkingPicker, convert all tensors to NumPy array and convert them back during deserialization. This is to decrease the consumption of file descriptors in the dataloader.

Here is the speed comparison in MobileNetV2 (batch size 64)

	inference time	FPS
Original	19.5 ms	3289
New	14.1 ms	4531

BC-breaking (Optional)

The Mixup, CutMix, and ResizeMix classes won't accept the num_classes argument, and it becomes the argument of ClsDataPreprocessor.

In the config files, the changes are as below:

# --------- Original config --------
model = dict(
    ...
    train_cfg=dict(augments=[
        dict(type='Mixup', alpha=0.2, num_classes=1000),
        dict(type='CutMix', alpha=1.0, num_classes=1000)
    ]),
)

data_preprocessor = dict(
    mean=[123.675, 116.28, 103.53],
    std=[58.395, 57.12, 57.375],
    to_rgb=True,
)

# -------- New config --------
model = dict(
    ...
    train_cfg=dict(augments=[
        dict(type='Mixup', alpha=0.2,
        dict(type='CutMix', alpha=1.0)
    ]),
)

data_preprocessor = dict(
    num_classes=1000,
    mean=[123.675, 116.28, 103.53],
    std=[58.395, 57.12, 57.375],
    to_rgb=True,
)

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here and update the documentation.

Checklist

Before PR:

• Pre-commit or other linting tools are used to fix the potential lint issues.
• Bug fixes are fully covered by unit tests, the case that causes the bug should be added in the unit tests.
• The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
• The documentation has been modified accordingly, like docstring or example tutorials.

After PR:

• If the modification has potential influence on downstream or other related projects, this PR should be tested with those projects, like MMDet or MMSeg.
• CLA has been signed and all committers have signed the CLA in this PR.

[codecov]

Codecov Report

Base: 0.02% // Head: 91.32% // Increases project coverage by +91.29% 🎉

Coverage data is based on head (08dd17c) compared to base (b8b31e9).
Patch has no changes to coverable lines.

❗ Current head 08dd17c differs from pull request most recent head 833ebe7. Consider uploading reports for the commit 833ebe7 to get more accurate results

Additional details and impacted files

@@             Coverage Diff              @@
##           dev-1.x    #1064       +/-   ##
============================================
+ Coverage     0.02%   91.32%   +91.29%     
============================================
  Files          121      128        +7     
  Lines         8217     9509     +1292     
  Branches      1368     1498      +130     
============================================
+ Hits             2     8684     +8682     
+ Misses        8215      639     -7576     
- Partials         0      186      +186     

Flag	Coverage Δ
unittests	91.32% <ø> (+91.29%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmcls/apis/inference.py	0.00% <0.00%> (ø)
mmcls/datasets/transforms/compose.py
mmcls/models/backbones/swin_transformer_v2.py	89.47% <0.00%> (ø)
mmcls/models/backbones/efficientformer.py	95.08% <0.00%> (ø)
mmcls/models/heads/efficientformer_head.py	93.10% <0.00%> (ø)
mmcls/models/backbones/edgenext.py	95.20% <0.00%> (ø)
mmcls/models/utils/layer_scale.py	86.66% <0.00%> (ø)
mmcls/models/backbones/mvit.py	92.46% <0.00%> (ø)
mmcls/models/backbones/mobileone.py	94.47% <0.00%> (ø)
mmcls/structures/utils.py	77.77% <0.00%> (ø)
... and 118 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[tonysy]

This line can be removed to keep consistent with other config file

[tonysy]

To help the user for understanding our codebase, maybe we need to update the document by adding some description of the data_samples used in MMClassification(with text or link to MMEngine related document), this could be added into the document TODO list.

[tonysy]

We can create an issue and add it to Oct. TODO list

[tonysy]

LGTM