regressions-msmarco-passage-cohere-embed-english-v3.0-hnsw-int8.md

Anserini Regressions: MS MARCO Passage Ranking

Model: Cohere embed-english-v3.0 with HNSW quantized indexes (using pre-encoded queries)

This page describes regression experiments, integrated into Anserini's regression testing framework, using the Cohere embed-english-v3.0 model on the MS MARCO passage ranking task.

In these experiments, we are using pre-encoded queries (i.e., cached results of query encoding).

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-passage-cohere-embed-english-v3.0-hnsw-int8

We make available a version of the MS MARCO Passage Corpus that has already been encoded with Cohere embed-english-v3.0.

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-passage-cohere-embed-english-v3.0-hnsw-int8

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-passage-cohere-embed-english-v3.0.tar -P collections/
tar xvf collections/msmarco-passage-cohere-embed-english-v3.0.tar -C collections/

To confirm, msmarco-passage-cohere-embed-english-v3.0.tar is 38 GB and has MD5 checksum 06a6e38a0522850c6aa504db7b2617f5. 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-passage-cohere-embed-english-v3.0-hnsw-int8 \
  --corpus-path collections/msmarco-passage-cohere-embed-english-v3.0

Indexing

Sample indexing command, building HNSW indexes:

target/appassembler/bin/IndexHnswDenseVectors \
  -collection JsonDenseVectorCollection \
  -input /path/to/msmarco-passage-cohere-embed-english-v3.0 \
  -generator HnswDenseVectorDocumentGenerator \
  -index indexes/lucene-hnsw.msmarco-passage-cohere-embed-english-v3.0-int8/ \
  -threads 16 -M 16 -efC 100 -noMerge -quantize.int8 \
  >& logs/log.msmarco-passage-cohere-embed-english-v3.0 &

The path /path/to/msmarco-passage-cohere-embed-english-v3.0/ should point to the corpus downloaded above. Upon completion, we should have an index with 8,841,823 documents.

Note that here we are explicitly using Lucene's NoMergePolicy merge policy, which suppresses any merging of index segments. This is because merging index segments is a costly operation and not worthwhile given our query set. Furthermore, we are using Lucene's Automatic Byte Quantization feature, which increase the on-disk footprint of the indexes since we're storing both the int8 quantized vectors and the float32 vectors, but only the int8 quantized vectors need to be loaded into memory. See issue #2292 for some experiments reporting the performance impact.

Retrieval

Topics and qrels are stored here, which is linked to the Anserini repo as a submodule. The regression experiments here evaluate on the 6980 dev set questions; see this page for more details.

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

target/appassembler/bin/SearchHnswDenseVectors \
  -index indexes/lucene-hnsw.msmarco-passage-cohere-embed-english-v3.0-int8/ \
  -topics tools/topics-and-qrels/topics.msmarco-passage.dev-subset.cohere-embed-english-v3.0.jsonl.gz \
  -topicReader JsonIntVector \
  -output runs/run.msmarco-passage-cohere-embed-english-v3.0.cohere-embed-english-v3.0.topics.msmarco-passage.dev-subset.cohere-embed-english-v3.0.jsonl.txt \
  -generator VectorQueryGenerator -topicField vector -threads 16 -hits 1000 -efSearch 1000 &

Evaluation can be performed using trec_eval:

target/appassembler/bin/trec_eval -c -m ndcg_cut.10 tools/topics-and-qrels/qrels.msmarco-passage.dev-subset.txt runs/run.msmarco-passage-cohere-embed-english-v3.0.cohere-embed-english-v3.0.topics.msmarco-passage.dev-subset.cohere-embed-english-v3.0.jsonl.txt
target/appassembler/bin/trec_eval -c -m map tools/topics-and-qrels/qrels.msmarco-passage.dev-subset.txt runs/run.msmarco-passage-cohere-embed-english-v3.0.cohere-embed-english-v3.0.topics.msmarco-passage.dev-subset.cohere-embed-english-v3.0.jsonl.txt
target/appassembler/bin/trec_eval -c -M 10 -m recip_rank tools/topics-and-qrels/qrels.msmarco-passage.dev-subset.txt runs/run.msmarco-passage-cohere-embed-english-v3.0.cohere-embed-english-v3.0.topics.msmarco-passage.dev-subset.cohere-embed-english-v3.0.jsonl.txt
target/appassembler/bin/trec_eval -c -m recall.1000 tools/topics-and-qrels/qrels.msmarco-passage.dev-subset.txt runs/run.msmarco-passage-cohere-embed-english-v3.0.cohere-embed-english-v3.0.topics.msmarco-passage.dev-subset.cohere-embed-english-v3.0.jsonl.txt

Effectiveness

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

nDCG@10	cohere-embed-english-v3.0
MS MARCO Passage: Dev	0.427
AP@1000	cohere-embed-english-v3.0
MS MARCO Passage: Dev	0.371
RR@10	cohere-embed-english-v3.0
MS MARCO Passage: Dev	0.365
R@1000	cohere-embed-english-v3.0
MS MARCO Passage: Dev	0.974

Note that due to the non-deterministic nature of HNSW indexing, results may differ slightly between each experimental run. Nevertheless, scores are generally within 0.005 of the reference values recorded in our YAML configuration file.

Reproduction Log*

To add to this reproduction log, modify this template and run bin/build.sh to rebuild the documentation.