Skip to content

Latest commit

 

History

History
120 lines (88 loc) · 7.82 KB

regressions-msmarco-v2-passage.splade-pp-sd.onnx.md

File metadata and controls

120 lines (88 loc) · 7.82 KB
Raw

Anserini Regressions: MS MARCO (V2) Passage Ranking

Model: SPLADE++ CoCondenser-SelfDistil (using ONNX for on-the-fly query encoding)

This page describes regression experiments, integrated into Anserini's regression testing framework, applying the SPLADE++ CoCondenser-SelfDistil model to the MS MARCO V2 passage corpus. Here, we evaluate on the dev queries, using ONNX to perform query encoding on the fly.

The model is described in the following paper:

Thibault Formal, Carlos Lassance, Benjamin Piwowarski, and Stéphane Clinchant. From Distillation to Hard Negative Sampling: Making Sparse Neural IR Models More Effective. Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 2353–2359.

For additional instructions on working with the MS MARCO V2 passage corpus, refer to this page.

The exact configurations for these regressions are stored in this YAML file. Note that this page is automatically generated from this template as part of Anserini's regression pipeline, so do not modify this page directly; modify the template instead and then run bin/build.sh to rebuild the documentation.

From one of our Waterloo servers (e.g., orca), the following command will perform the complete regression, end to end:

python src/main/python/run_regression.py --index --verify --search --regression msmarco-v2-passage.splade-pp-sd.onnx

We make available a version of the corpus that has already been encoded with SPLADE++ CoCondenser-SelfDistil.

From any machine, the following command will download the corpus and perform the complete regression, end to end:

python src/main/python/run_regression.py --download --index --verify --search --regression msmarco-v2-passage.splade-pp-sd.onnx

The run_regression.py script automates the following steps, but if you want to perform each step manually, simply copy/paste from the commands below and you'll obtain the same regression results.

Corpus Download

Download the corpus and unpack into collections/:

wget https://rgw.cs.uwaterloo.ca/pyserini/data/msmarco_v2_passage_splade_pp_sd.tar -P collections/
tar xvf collections/msmarco_v2_passage_splade_pp_sd.tar -C collections/

To confirm, msmarco_v2_passage_splade_pp_sd.tar is 76 GB and has MD5 checksum 061930dd615c7c807323ea7fc7957877. With the corpus downloaded, the following command will perform the remaining steps below:

python src/main/python/run_regression.py --index --verify --search --regression msmarco-v2-passage.splade-pp-sd.onnx \
  --corpus-path collections/msmarco_v2_passage_splade_pp_sd

Indexing

Sample indexing command:

bin/run.sh io.anserini.index.IndexCollection \
  -threads 24 \
  -collection JsonVectorCollection \
  -input /path/to/msmarco-v2-passage-splade-pp-sd \
  -generator DefaultLuceneDocumentGenerator \
  -index indexes/lucene-inverted.msmarco-v2-passage.splade-pp-sd/ \
  -impact -pretokenized -storeRaw \
  >& logs/log.msmarco-v2-passage-splade-pp-sd &

The path /path/to/msmarco-v2-passage-splade-pp-sd/ should point to the corpus downloaded above.

The important indexing options to note here are -impact -pretokenized: the first tells Anserini not to encode BM25 doc lengths into Lucene's norms (which is the default) and the second option says not to apply any additional tokenization on the pre-encoded tokens. Upon completion, we should have an index with 8,841,823 documents.

For additional details, see explanation of common indexing options.

Retrieval

Topics and qrels are stored here, which is linked to the Anserini repo as a submodule.

After indexing has completed, you should be able to perform retrieval as follows:

bin/run.sh io.anserini.search.SearchCollection \
  -index indexes/lucene-inverted.msmarco-v2-passage.splade-pp-sd/ \
  -topics tools/topics-and-qrels/topics.msmarco-v2-passage.dev.txt \
  -topicReader TsvInt \
  -output runs/run.msmarco-v2-passage-splade-pp-sd.splade-pp-sd-onnx.topics.msmarco-v2-passage.dev.txt \
  -parallelism 16 -impact -pretokenized -encoder SpladePlusPlusSelfDistil &
bin/run.sh io.anserini.search.SearchCollection \
  -index indexes/lucene-inverted.msmarco-v2-passage.splade-pp-sd/ \
  -topics tools/topics-and-qrels/topics.msmarco-v2-passage.dev2.txt \
  -topicReader TsvInt \
  -output runs/run.msmarco-v2-passage-splade-pp-sd.splade-pp-sd-onnx.topics.msmarco-v2-passage.dev2.txt \
  -parallelism 16 -impact -pretokenized -encoder SpladePlusPlusSelfDistil &

Evaluation can be performed using trec_eval:

bin/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.msmarco-v2-passage.dev.txt runs/run.msmarco-v2-passage-splade-pp-sd.splade-pp-sd-onnx.topics.msmarco-v2-passage.dev.txt
bin/trec_eval -c -m recall.1000 tools/topics-and-qrels/qrels.msmarco-v2-passage.dev.txt runs/run.msmarco-v2-passage-splade-pp-sd.splade-pp-sd-onnx.topics.msmarco-v2-passage.dev.txt
bin/trec_eval -c -M 100 -m map -c -M 100 -m recip_rank tools/topics-and-qrels/qrels.msmarco-v2-passage.dev.txt runs/run.msmarco-v2-passage-splade-pp-sd.splade-pp-sd-onnx.topics.msmarco-v2-passage.dev.txt
bin/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.msmarco-v2-passage.dev2.txt runs/run.msmarco-v2-passage-splade-pp-sd.splade-pp-sd-onnx.topics.msmarco-v2-passage.dev2.txt
bin/trec_eval -c -m recall.1000 tools/topics-and-qrels/qrels.msmarco-v2-passage.dev2.txt runs/run.msmarco-v2-passage-splade-pp-sd.splade-pp-sd-onnx.topics.msmarco-v2-passage.dev2.txt
bin/trec_eval -c -M 100 -m map -c -M 100 -m recip_rank tools/topics-and-qrels/qrels.msmarco-v2-passage.dev2.txt runs/run.msmarco-v2-passage-splade-pp-sd.splade-pp-sd-onnx.topics.msmarco-v2-passage.dev2.txt

Effectiveness

With the above commands, you should be able to reproduce the following results:

MAP@100 SPLADE++ CoCondenser-SelfDistil
MS MARCO V2 Passage: Dev 0.1522
MS MARCO V2 Passage: Dev2 0.1613
MRR@100 SPLADE++ CoCondenser-SelfDistil
MS MARCO V2 Passage: Dev 0.1536
MS MARCO V2 Passage: Dev2 0.1627
R@100 SPLADE++ CoCondenser-SelfDistil
MS MARCO V2 Passage: Dev 0.6085
MS MARCO V2 Passage: Dev2 0.6178
R@1000 SPLADE++ CoCondenser-SelfDistil
MS MARCO V2 Passage: Dev 0.8268
MS MARCO V2 Passage: Dev2 0.8234