Completed replication for msmarco-doc and msmarco-passage

[shaneding]

Environment
OS: macOS 10.15 (Catalina)
Java: 14.0.1
Python: 3.7.6

Issues:
BM25 Baseline on MS Marco Passage Retreival:
Error: return self.object.search(JString(q.encode('utf8')), k)
AttributeError: 'bytes' object has no attribute 'encode’

• Encountered when using pyserini to retrieve the smaller set of queries, had to use java implementation instead.

BM25 Baselines on MS MARCO Doc Retrieval Task

• Within replication doc (for comparison to Microsoft baseline), the command given was: eval/trec_eval.9.0.4/trec_eval -c -mmap -M 100 collections/msmarco-doc/queries-and-qrels/msmarco-docdev-qrels.tsv collections/msmarco-doc/msmarco-docdev-top100.
• I had to modify this to: eval/trec_eval.9.0.4/trec_eval -c -mmap -M 100 collections/msmarco-doc/queries-and-qrels/msmarco-docdev-qrels.tsv collections/msmarco-doc/queries-and-qrels/msmarco-docdev-top100
• Key difference was that msmarco-docdev-top100 was stored in queries-and-qrels rather than directly msmarco-doc

[codecov]

Codecov Report

Merging #1220 into master will increase coverage by 0.01%.
The diff coverage is n/a.

@@             Coverage Diff              @@
##             master    #1220      +/-   ##
============================================
+ Coverage     48.34%   48.35%   +0.01%     
- Complexity      738      739       +1     
============================================
  Files           147      147              
  Lines          8547     8547              
  Branches       1212     1212              
============================================
+ Hits           4132     4133       +1     
+ Misses         4075     4074       -1     
  Partials        340      340              

Impacted Files	Coverage Δ	Complexity Δ
...anserini/ltr/feature/base/PMIFeatureExtractor.java	84.61% <0.00%> (-1.93%)	12.00% <0.00%> (-1.00%)
...java/io/anserini/ltr/feature/CountBigramPairs.java	89.61% <0.00%> (+2.59%)	33.00% <0.00%> (+2.00%)

---

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 b6e0367...19524a3. Read the comment docs.

[lintool]

Can you update the MS MARCO doc to directly fix the issue you encountered. Probably leftover from a recent refactoring...

[kelvin-jiang]

Can confirm that I'm running into the same AttributeError: 'bytes' object has no attribute 'encode’ when retrieving with Python. Should I file an issue in the pyserini repo?

[lintool]

@kelvin-jiang yes please. the issue should be filed in this repo, though.

I think it might be because PyPI is behind the latest HEAD on master, lets record in issue regardless.