Parameters of DilatedEncoder

[Shanyaliux]

Add the original fixed parameter block_dilations as an incoming parameter

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily get feedback. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.

Motivation

When I used yolof, I modified the modification and found that the parameter named block_dilations of DilatedEncoder is fixed. If I have num_residual_blocks as other values, this parameter cannot be very convenient.

Modification

I modified this parameter of DilatedEncoder to be an incoming parameter and added a default value in its config file.

• DilatedEncoder modifies the position
  https://github.com/Shanyaliux/mmdetection/blob/aa99c91d64f4de35ef8040493d0cf5453d662476/mmdet/models/necks/dilated_encoder.py#L68
  https://github.com/Shanyaliux/mmdetection/blob/aa99c91d64f4de35ef8040493d0cf5453d662476/mmdet/models/necks/dilated_encoder.py#L74
• yolof configuration file modification location
  https://github.com/Shanyaliux/mmdetection/blob/aa99c91d64f4de35ef8040493d0cf5453d662476/configs/yolof/yolof_r50_c5_8x8_1x_coco.py#L25

BC-breaking (Optional)

Does the modification introduce changes that break the backward-compatibility of the downstream repos?
If so, please describe how it breaks the compatibility and how the downstream projects should modify their code to keep compatibility with this PR.

Use cases (Optional)

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

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
3. If the modification has potential influence on downstream projects, this PR should be tested with downstream projects, like MMDet or MMCls.
4. The documentation has been modified accordingly, like docstring or example tutorials.

[CLAassistant]

All committers have signed the CLA.

[Shanyaliux]

In the test process after the first submission, I noticed that the error was reported because the new block_dilations was missing when its test function called DilatedEncoder.
The error message is as follows:

    def test_dilated_encoder():
        in_channels = 16
        out_channels = 32
        out_shape = 34
>       dilated_encoder = DilatedEncoder(in_channels, out_channels, 16, 2)
E       TypeError: __init__() missing 1 required positional argument: 'block_dilations'

So I fixed the bug in the latest commit

[Shanyaliux]

I noticed that you released v2.24.0, do I need to resubmit on the new version?

[codecov]

Codecov Report

Merging #7812 (e17bae3) into dev (73b4e65) will decrease coverage by 0.03%.
The diff coverage is 100.00%.

❗ Current head e17bae3 differs from pull request most recent head 942701e. Consider uploading reports for the commit 942701e to get more accurate results

@@            Coverage Diff             @@
##              dev    #7812      +/-   ##
==========================================
- Coverage   64.71%   64.67%   -0.04%     
==========================================
  Files         351      351              
  Lines       28461    28461              
  Branches     4807     4807              
==========================================
- Hits        18418    18408      -10     
- Misses       9055     9066      +11     
+ Partials      988      987       -1     

Flag	Coverage Δ
unittests	64.67% <100.00%> (-0.04%)	⬇️

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

Impacted Files	Coverage Δ
mmdet/models/necks/dilated_encoder.py	82.35% <100.00%> (ø)
mmdet/core/bbox/samplers/random_sampler.py	75.00% <0.00%> (-5.56%)	⬇️
mmdet/models/detectors/cornernet.py	94.87% <0.00%> (-5.13%)	⬇️
mmdet/core/bbox/samplers/sampling_result.py	73.01% <0.00%> (-1.59%)	⬇️
mmdet/models/dense_heads/corner_head.py	68.33% <0.00%> (-1.39%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 73b4e65...942701e. Read the comment docs.

[Shanyaliux]

Sorry, I don't know how to solve this error, it shouldn't be my problem.

[ZwwWayne]

Sorry, I don't know how to solve this error, it shouldn't be my problem.

Do not worry, it is caused by other issues and have been fixed in dev branch.

[ZwwWayne]

Sorry, I don't know how to solve this error, it shouldn't be my problem.

Do not worry, it is caused by other issues and have been fixed in dev branch.