Model output test

[sgugger]

Step 2 of the strategy for the new model outputs as outlined on the forum.
Use the return_dict argument introduced in #6138 in all tests and remove all unpacking from the tests.

[codecov]

Codecov Report

Merging #6155 into master will increase coverage by 0.10%.
The diff coverage is 75.00%.

@@            Coverage Diff             @@
##           master    #6155      +/-   ##
==========================================
+ Coverage   78.35%   78.46%   +0.10%     
==========================================
  Files         146      146              
  Lines       26454    26454              
==========================================
+ Hits        20729    20758      +29     
+ Misses       5725     5696      -29     

Impacted Files	Coverage Δ
src/transformers/modeling_encoder_decoder.py	92.20% <ø> (ø)
src/transformers/modeling_openai.py	82.25% <66.66%> (-0.38%)	⬇️
src/transformers/modeling_reformer.py	95.49% <100.00%> (-0.21%)	⬇️
src/transformers/modeling_tf_flaubert.py	24.22% <0.00%> (-63.98%)	⬇️
src/transformers/modeling_longformer.py	89.21% <0.00%> (-2.24%)	⬇️
src/transformers/modeling_electra.py	81.55% <0.00%> (-0.29%)	⬇️
src/transformers/modeling_t5.py	83.71% <0.00%> (ø)
src/transformers/modeling_xlnet.py	81.75% <0.00%> (ø)
... and 5 more

---

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 91cb954...33ebdb9. Read the comment docs.

[LysandreJik]

Impressive work, LGTM!