[Fix] fix_torchserver1.1

[VVsssssk]

Hi,when I try to test mmseg's torchserver function,i find some problems that casue server can't run.So,I fix problem and write a python script to test function.

• 1,when I use cpu env to inference a fcn model

2021-08-31 07:34:25,801 [INFO ] W-9006-fcn_1.0-stdout MODEL_LOG - raise ValueError('SyncBatchNorm expected input tensor to be on GPU')
2021-08-31 07:34:25,801 [INFO ] W-9006-fcn_1.0-stdout MODEL_LOG - ValueError: SyncBatchNorm expected input tensor to be on GPU

• 2,in postprocess output result to json

2021-08-31 07:58:56,594 [INFO ] W-9000-fcn_1.0-stdout MODEL_LOG - Unable to serialize model output.
2021-08-31 07:58:56,595 [INFO ] W-9000-fcn_1.0-stdout MODEL_LOG - raise TypeError(f'Object of type {o.class.name} '
2021-08-31 07:58:56,595 [INFO ] W-9000-fcn_1.0-stdout MODEL_LOG - TypeError: Object of type ndarray is not JSON serializable

[CLAassistant]

All committers have signed the CLA.

[Junjun2016]

Hi @VVsssssk
Thanks for your contribution.
We can import synbn2bn from mmcv after it merged.
Please refer to #772.

[codecov]

Codecov Report

Merging #844 (ba0747f) into master (4e6f2eb) will decrease coverage by 0.00%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #844      +/-   ##
==========================================
- Coverage   89.05%   89.05%   -0.01%     
==========================================
  Files         111      112       +1     
  Lines        6051     6060       +9     
  Branches      969      970       +1     
==========================================
+ Hits         5389     5397       +8     
- Misses        467      468       +1     
  Partials      195      195              

Flag	Coverage Δ
unittests	89.05% <ø> (-0.01%)	⬇️

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

Impacted Files	Coverage Δ
mmseg/datasets/__init__.py	100.00% <0.00%> (ø)
mmseg/datasets/coco_stuff.py	87.50% <0.00%> (ø)

---

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 4e6f2eb...ba0747f. Read the comment docs.

[RunningLeon]

@VVsssssk Could fix conflicts at first.

[Junjun2016]

@zhouzaida
When will syncbn2bn merge in mmcv?
And maybe mmcv should also support bn2syncbn.

[Junjun2016]

Can import revert_sync_batchnorm from open-mmlab/mmcv#1253

[RunningLeon]

Can import revert_sync_batchnorm from open-mmlab/mmcv#1253

@VVsssssk Hi, may update accordingly.

[Junjun2016]

Hi @VVsssssk
Could you please update this PR.

[Junjun2016]

Please resolve the comments.

[Junjun2016]

Please also resolve the conflict.

[Junjun2016]

LGTM

[Junjun2016]

Hi @RunningLeon
Please review it again.

[RunningLeon]

LGTM