Run GPU tests in individual subprocesses

[leezu]

Description

Following the TVM PR, CI is failing with out of memory issues. Test skipping TVM tests.

Reference of CI running out of memory even with an older MXNet commit: https://github.com/dmlc/gluon-nlp/actions/runs/328560026

[codecov[bot]]

Codecov Report

Merging #1409 into master will increase coverage by 0.25%. The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #1409      +/-   ##
==========================================
+ Coverage   85.12%   85.38%   +0.25%     
==========================================
  Files          53       53              
  Lines        6946     6946              
==========================================
+ Hits         5913     5931      +18     
+ Misses       1033     1015      -18     

Impacted Files	Coverage Δ
src/gluonnlp/attention_cell.py	80.31% <0.00%> (-0.40%)	:arrow_down:
src/gluonnlp/data/filtering.py	82.60% <0.00%> (+4.34%)	:arrow_up:
src/gluonnlp/data/loading.py	83.39% <0.00%> (+5.28%)	:arrow_up:

---

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 7c1927f...ad5d1c9. Read the comment docs.

[leezu]

CI passed with the "old" MXNet commit. @barry-jin helped to update the github container and I'll now retrigger this PR to see if it still "fixes" the CI issue with the latest MXNet version

[leezu]

Fails with the updated MXNet commit due to out of gpu memory

[2020-10-29T00:29:33.285Z] tests/test_models.py ............................ssssssssssssssss        [ 57%]
[2020-10-29T00:31:03.325Z] tests/test_models_albert.py .................                            [ 58%]
[2020-10-29T00:32:20.559Z] tests/test_models_bart.py ......                                         [ 59%]
[2020-10-29T00:39:07.063Z] tests/test_models_bert.py ............                                   [ 60%]
[2020-10-29T00:41:36.402Z] tests/test_models_electra.py ........                                    [ 60%]
[2020-10-29T00:46:45.319Z] tests/test_models_gpt2.py .......F                                       [ 61%]
[2020-10-29T00:46:57.180Z] tests/test_models_mobilebert.py .....                                    [ 61%]
[2020-10-29T00:49:06.186Z] tests/test_models_roberta.py ....FF                                      [ 62%]
[2020-10-29T00:49:36.862Z] tests/test_models_transformer.py ....................................... [ 65%]
[2020-10-29T00:50:37.228Z] ........................................................................ [ 71%]
[2020-10-29T00:51:18.688Z] ..........................................FFFFF                          [ 74%]
[2020-10-29T00:51:23.502Z] tests/test_models_transformer_xl.py ......                               [ 75%]
[2020-10-29T00:53:05.532Z] tests/test_models_xlmr.py .FF                                            [ 75%]
[2020-10-29T00:53:05.752Z] tests/test_op.py ....................................................... [ 79%]
[2020-10-29T00:53:06.194Z] ........................................................................ [ 85%]
[2020-10-29T00:53:06.747Z] ....                                                                     [ 85%]
[2020-10-29T00:53:38.587Z] tests/test_optimizer.py .                                                [ 85%]
[2020-10-29T00:53:38.592Z] tests/test_pytest.py .                                                   [ 85%]
[2020-10-29T00:53:38.741Z] tests/test_sequence_sampler.py ......................................... [ 89%]
[2020-10-29T00:53:39.027Z] ........................................................................ [ 94%]
[2020-10-29T00:53:45.583Z] .......................................                                  [ 97%]

[github-actions[bot]]

The documentation website for preview: http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR1409/leezu-patch-1/index.html

[github-actions[bot]]

The documentation website for preview: http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR1409/leezu-patch-1/index.html

[github-actions[bot]]

The documentation website for preview: http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR1409/leezu-patch-1/index.html

[leezu]

Thanks to @barry-jin to point out that the status displayed in this PR does not take into account the changes in the yml files.

Please refer to https://github.com/dmlc/gluon-nlp/actions/runs/334873655/workflow for the passing tests.

Let's merge this PR to unblock the CI @szha