[s2s] distributed eval allows num_return_sequences > 1

[sshleifer]

For more complicated pseudolabeling procedures.

[codecov]

Codecov Report

Merging #7254 into master will increase coverage by 1.08%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #7254      +/-   ##
==========================================
+ Coverage   78.20%   79.29%   +1.08%     
==========================================
  Files         181      181              
  Lines       35751    35751              
==========================================
+ Hits        27959    28347     +388     
+ Misses       7792     7404     -388     

Impacted Files	Coverage Δ
src/transformers/modeling_tf_longformer.py	17.16% <0.00%> (-81.42%)	⬇️
src/transformers/modeling_layoutlm.py	25.06% <0.00%> (-69.40%)	⬇️
src/transformers/configuration_layoutlm.py	80.00% <0.00%> (-20.00%)	⬇️
src/transformers/modeling_roberta.py	81.13% <0.00%> (-15.42%)	⬇️
src/transformers/modeling_mobilebert.py	79.22% <0.00%> (-10.17%)	⬇️
src/transformers/modeling_bert.py	88.36% <0.00%> (+0.16%)	⬆️
src/transformers/tokenization_utils_base.py	93.91% <0.00%> (+0.27%)	⬆️
src/transformers/modeling_bart.py	94.44% <0.00%> (+0.50%)	⬆️
src/transformers/generation_tf_utils.py	86.46% <0.00%> (+1.25%)	⬆️
src/transformers/modeling_auto.py	84.03% <0.00%> (+1.40%)	⬆️
... and 8 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 0ccb6f5...2960c40. Read the comment docs.