SPLADE++ ED w/ ONNX on BEIR

[lintool]

Here's the snippet for running everything from scratch:

MODEL="splade-pp-ed-onnx"; CORPORA=(trec-covid bioasq nfcorpus nq hotpotqa fiqa signal1m trec-news robust04 arguana webis-touche2020 cqadupstack-android cqadupstack-english cqadupstack-gaming cqadupstack-gis cqadupstack-mathematica cqadupstack-physics cqadupstack-programmers cqadupstack-stats cqadupstack-tex cqadupstack-unix cqadupstack-webmasters cqadupstack-wordpress quora dbpedia-entity scidocs fever climate-fever scifact); for c in "${CORPORA[@]}"
do
    echo "Running $c..."
    python src/main/python/run_regression.py --index --verify --search --regression beir-v1.0.0-${c}-${MODEL} > logs/log.beir-v1.0.0-${c}-${MODEL} 2>&1
done

Everything works except for:

python src/main/python/run_regression.py --index --search --regression beir-v1.0.0-arguana-splade-pp-ed-onnx

@ArthurChen189 can you please confirm repo of error? and then try to fix?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (f86a65f) 65.66% compared to head (2fd6544) 65.66%.

Additional details and impacted files

@@            Coverage Diff            @@
##             master    #2354   +/-   ##
=========================================
  Coverage     65.66%   65.66%           
  Complexity     1398     1398           
=========================================
  Files           207      207           
  Lines         11631    11631           
  Branches       1471     1471           
=========================================
  Hits           7637     7637           
  Misses         3484     3484           
  Partials        510      510           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[lintool]

Everything works except for arguana (still), which @ArthurChen189 is looking into. I'm going to merge in this PR since we need it regardless; @ArthurChen189 can work on fixing the bug in parallel.

[carlos-lassance]

@ArthurChen189 arguana could be a max length on the query side? We used 256 on both query and doc from what I remember.