[FEATURE] T5NMTInference model and incremental decoding

[yongyi-wu]

Description

This PR contains a T5NMTInference model and the incremental decoding feature. Test codes are also included. Overall, the design follows the implementations in transformer.py, so it should be straightforward for future maintenance.

Checklist

Essentials

• [x] PR's title starts with a category (e.g. [BUGFIX], [MODEL], [TUTORIAL], [FEATURE], [DOC], etc)
• [x] Changes are complete (i.e. I finished coding on this PR)
• [x] All changes have test coverage

Changes

• [x] T5NMTInference model
• [x] incremental_decoding()
• [x] refactored relative position encoding

cc @dmlc/gluon-nlp-team

[github-actions[bot]]

The documentation website for preview: http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR1498/f10e9749355597a57bcca88d3b9b1cd344e03f87/index.html

[codecov[bot]]

Codecov Report

Merging #1498 (ee8b749) into master (36092e2) will increase coverage by 0.75%. The diff coverage is 98.86%.

@@            Coverage Diff             @@
##           master    #1498      +/-   ##
==========================================
+ Coverage   85.84%   86.60%   +0.75%     
==========================================
  Files          53       53              
  Lines        7265     7323      +58     
==========================================
+ Hits         6237     6342     +105     
+ Misses       1028      981      -47     

Impacted Files	Coverage Δ
src/gluonnlp/utils/testing.py	94.16% <ø> (ø)
src/gluonnlp/models/t5.py	98.13% <98.71%> (+6.39%)	:arrow_up:
src/gluonnlp/attention_cell.py	88.39% <100.00%> (+0.54%)	:arrow_up:
src/gluonnlp/data/tokenizers/huggingface.py	71.54% <0.00%> (-0.29%)	:arrow_down:
src/gluonnlp/layers.py	86.72% <0.00%> (+0.84%)	:arrow_up:
src/gluonnlp/data/filtering.py	82.60% <0.00%> (+4.34%)	:arrow_up:
src/gluonnlp/data/loading.py	83.39% <0.00%> (+7.54%)	: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 36092e2...ee8b749. Read the comment docs.

[github-actions[bot]]

The documentation website for preview: http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR1498/ee8b749335cdd9e2474dc010c8220a984831d17f/index.html