Merge branch 'master' into add-fusion-regression

castorini · Nov 25, 2024 · 5e8cea1 · 5e8cea1
2 parents 149b298 + a5e6771
commit 5e8cea1
Show file tree

Hide file tree

Showing 1,123 changed files with 49,993 additions and 5,150 deletions.
diff --git a/README.md b/README.md
diff --git a/docs/experiments-msmarco-passage.md b/docs/experiments-msmarco-passage.md
@@ -25,7 +25,7 @@ Anserini is a toolkit (in Java) for reproducible information retrieval research
 The Lucene search library provides components of the popular [Elasticsearch](https://www.elastic.co/) platform.
 
 Think of it this way: Lucene provides a "kit of parts".
-Elasticsearch provides "assembly of parts" targeted to production search applications, with a REST-centric API.
+Elasticsearch provides an "assembly of parts" targeted to production search applications, with a REST-centric API.
 Anserini provides an alternative way of composing the same core components together, targeted at information retrieval researchers.
 By building on Lucene, Anserini aims to bridge the gap between academic information retrieval research and the practice of building real-world search applications.
 That is, most things done with Anserini can be "translated" into Elasticsearch quite easily.
@@ -100,7 +100,7 @@ bin/run.sh io.anserini.index.IndexCollection \
   -input collections/msmarco-passage/collection_jsonl \
   -index indexes/msmarco-passage/lucene-index-msmarco \
   -generator DefaultLuceneDocumentGenerator \
-  -threads 9 -storePositions -storeDocvectors -storeRaw 
+  -threads 9 -storePositions -storeDocvectors -storeRaw
 ```
 For Windows:
 ```bash
@@ -206,7 +206,7 @@ Since the first column indicates the `qid`, it means that the file contains rank
 
 ## Evaluation
 
-Finally, we can evaluate the retrieved documents using this the official MS MARCO evaluation script: 
+Finally, we can evaluate the retrieved documents using this the official MS MARCO evaluation script:
 
 ```bash
 python tools/scripts/msmarco/msmarco_passage_eval.py \
@@ -244,7 +244,7 @@ We take the average of the scores across all queries (6980 in this case), and we
 You can find this run on the [MS MARCO Passage Ranking Leaderboard](https://microsoft.github.io/MSMARCO-Passage-Ranking-Submissions/leaderboard/) as the entry named "BM25 (Lucene8, tuned)", dated 2019/06/26.
 So you've just reproduced (part of) a leaderboard submission!
 
-We can also use the official [TREC](https://trec.nist.gov/) evaluation tool, `trec_eval`, to compute other metrics than MRR@10. 
+We can also use the official [TREC](https://trec.nist.gov/) evaluation tool, `trec_eval`, to compute other metrics than MRR@10.
 For that we first need to convert runs and qrels files to the TREC format:
 
 ```bash
@@ -523,3 +523,18 @@ The BM25 run with default parameters `k1=0.9`, `b=0.4` roughly corresponds to th
 + Results reproduced by [@antea-ab](https://github.com/antea-ab) on 2024-09-01 (commit [`e0a9578`](https://github.com/castorini/anserini/commit/e0a9578cd391674e8b3aa15ee25906b5fb442c9d))
 + Results reproduced by [@anshulsc](https://github.com/anshulsc) on 2024-09-06 (commit [`c096dff`](https://github.com/castorini/anserini/commit/c096dffe0d114af3bc4d8e4e71ebef4fe02bc94d))
 + Results reproduced by [@r-aya](https://github.com/r-aya) on 2024-09-07 (commit [`4319f89`](https://github.com/castorini/anserini/commit/4319f89472c4dd3359482f041dbcaee5202d8dd2))
++ Results reproduced by [@Amirkia1998](https://github.com/Amirkia1998) on 2024-09-20 (commit [`9e0cd5b`](https://github.com/castorini/anserini/commit/204bc87ef66e689773549ff804377eae20d5d7ce))
++ Results reproduced by [@CCarolD](https://github.com/CCarolD) on 2024-09-20 (commit [`2cb5d13`](https://github.com/castorini/anserini/commit/2cb5d1377862d49f70fa60cc452e96b31d815b94))
++ Results reproduced by [@pjyi2147](https://github.com/pjyi2147) on 2024-09-20 (commit [`2cb5d13`](https://github.com/castorini/anserini/commit/2cb5d1377862d49f70fa60cc452e96b31d815b94))
++ Results reproduced by [@krishh-p](https://github.com/krishh-p) on 2024-09-21 (commit [`f04321f`](https://github.com/castorini/anserini/commit/f04321f40b6eb64308ea90394749912b6199589d))
++ Results reproduced by [@andrewxucs](https://github.com/andrewxucs) on 2024-09-22 (commit [`4ef1de0`](https://github.com/castorini/anserini/commit/4ef1de032c98372298da63f11618ff0f0861b2a7))
++ Results reproduced by [@Hossein-Molaeian](https://github.com/Hossein-Molaeian) on 2024-09-22 (commit [`3641b48`](https://github.com/castorini/anserini/commit/3641b48688ced617a5a13c9029a174fccf0ef6c6))
++ Results reproduced by [@AhmedEssam19](https://github.com/AhmedEssam19) on 2024-09-27 (commit [`25523b4`](https://github.com/castorini/anserini/commit/25523b48597b4f061f7d016f888a124028b9b01f))
++ Results reproduced by [@sisixili](https://github.com/sisixili) on 2024-10-01 (commit [`25523b4`](https://github.com/castorini/anserini/commit/25523b48597b4f061f7d016f888a124028b9b01f))
++ Results reproduced by [@Raghav0005](https://github.com/Raghav0005) on 2024-10-07 (commit [`ee97c1d`](https://github.com/castorini/anserini/commit/ee97c1d5deeb684748723179711128f67557832c))
++ Results reproduced by [@a-y-m-a-n-c-h](https://github.com/a-y-m-a-n-c-h) on 2024-10-16 (commit [`0346842`](https://github.com/castorini/anserini/commit/03468423c820e1c0c38c9f48dc25d1f2f315831c))
++ Results reproduced by [@Samantha-Zhan](https://github.com/Samantha-Zhan) on 2024-10-20 (commit [`daceb40`](https://github.com/castorini/anserini/commit/daceb4084c8e8103e3e86c81a8e0d597d409220e))
++ Results reproduced by [@pxlin-09](https://github.com/pxlin-09) on 2024-10-26 (commit [`e2eb203`](https://github.com/castorini/anserini/commit/e2eb203b83dd643a356ee90f299c8877f6e108bd))
++ Results reproduced by [@b8zhong](https://github.com/b8zhong) on 2024-11-23 (commit [`c619dc8`](https://github.com/castorini/anserini/commit/c619dc8d9ab28298251964053a927906e9957f51))
++ Results reproduced by [@ShreyasP20](https://github.com/ShreyasP20) on 2024-11-24 (commit [`c619dc8`](https://github.com/castorini/anserini/commit/c619dc8d9ab28298251964053a927906e9957f51))
++ Results reproduced by [@Divyajyoti02](https://github.com/Divyajyoti02) on 2024-11-24 (commit [`a1bcf88`](https://github.com/castorini/anserini/commit/a1bcf8853062da9f73915c873968a4e998d4e904))
diff --git a/.../regressions/regressions-beir-v1.0.0-arguana.bge-base-en-v1.5.hnsw-int8.onnx.md b/.../regressions/regressions-beir-v1.0.0-arguana.bge-base-en-v1.5.hnsw-int8.onnx.md
@@ -38,7 +38,7 @@ bin/run.sh io.anserini.index.IndexHnswDenseVectors \
   -input /path/to/beir-v1.0.0-arguana.bge-base-en-v1.5 \
   -generator DenseVectorDocumentGenerator \
   -index indexes/lucene-hnsw-int8.beir-v1.0.0-arguana.bge-base-en-v1.5/ \
-  -M 16 -efC 100 -memoryBuffer 65536 -noMerge -quantize.int8 \
+  -M 16 -efC 100 -quantize.int8 \
   >& logs/log.beir-v1.0.0-arguana.bge-base-en-v1.5 &
 ```
 

diff --git a/docs/regressions/regressions-beir-v1.0.0-arguana.bge-base-en-v1.5.hnsw.onnx.md b/docs/regressions/regressions-beir-v1.0.0-arguana.bge-base-en-v1.5.hnsw.onnx.md
@@ -38,7 +38,7 @@ bin/run.sh io.anserini.index.IndexHnswDenseVectors \
   -input /path/to/beir-v1.0.0-arguana.bge-base-en-v1.5 \
   -generator DenseVectorDocumentGenerator \
   -index indexes/lucene-hnsw.beir-v1.0.0-arguana.bge-base-en-v1.5/ \
-  -M 16 -efC 100 -memoryBuffer 65536 -noMerge \
+  -M 16 -efC 100 \
   >& logs/log.beir-v1.0.0-arguana.bge-base-en-v1.5 &
 ```
 

diff --git a/...ns/regressions-beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.cached.md b/...ns/regressions-beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.cached.md
@@ -0,0 +1,84 @@
+# Anserini Regressions: BEIR (v1.0.0) &mdash; ArguAna
+
+**Model**: [BGE-base-en-v1.5](https://huggingface.co/BAAI/bge-base-en-v1.5) with quantized flat indexes (using cached queries)
+
+This page describes regression experiments, integrated into Anserini's regression testing framework, using the [BGE-base-en-v1.5](https://huggingface.co/BAAI/bge-base-en-v1.5) model on [BEIR (v1.0.0) &mdash; ArguAna](http://beir.ai/), as described in the following paper:
+
+> Shitao Xiao, Zheng Liu, Peitian Zhang, and Niklas Muennighoff. [C-Pack: Packaged Resources To Advance General Chinese Embedding.](https://arxiv.org/abs/2309.07597) _arXiv:2309.07597_, 2023.
+
+In these experiments, we are using cached queries (i.e., cached results of query encoding).
+
+The exact configurations for these regressions are stored in [this YAML file](../../src/main/resources/regression/beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.cached.yaml).
+Note that this page is automatically generated from [this template](../../src/main/resources/docgen/templates/beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.cached.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 beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.cached
+```
+
+All the BEIR corpora, encoded by the BGE-base-en-v1.5 model and stored in Parquet format, are available for download:
+
+```bash
+wget https://rgw.cs.uwaterloo.ca/pyserini/data/beir-v1.0.0-bge-base-en-v1.5.parquet.tar -P collections/
+tar xvf collections/beir-v1.0.0-bge-base-en-v1.5.parquet.tar -C collections/
+```
+
+The tarball is 194 GB and has MD5 checksum `c279f9fc2464574b482ec53efcc1c487`.
+After download and unpacking the corpora, the `run_regression.py` command above should work without any issue.
+
+## Indexing
+
+Sample indexing command, building quantized flat indexes:
+
+```
+bin/run.sh io.anserini.index.IndexFlatDenseVectors \
+  -threads 16 \
+  -collection ParquetDenseVectorCollection \
+  -input /path/to/beir-v1.0.0-arguana.bge-base-en-v1.5 \
+  -generator ParquetDenseVectorDocumentGenerator \
+  -index indexes/lucene-flat-int8.beir-v1.0.0-arguana.bge-base-en-v1.5/ \
+  -quantize.int8 \
+  >& logs/log.beir-v1.0.0-arguana.bge-base-en-v1.5 &
+```
+
+The path `/path/to/beir-v1.0.0-arguana.bge-base-en-v1.5/` should point to the corpus downloaded above.
+
+## Retrieval
+
+Topics and qrels are stored [here](https://github.com/castorini/anserini-tools/tree/master/topics-and-qrels), 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.SearchFlatDenseVectors \
+  -index indexes/lucene-flat-int8.beir-v1.0.0-arguana.bge-base-en-v1.5/ \
+  -topics tools/topics-and-qrels/topics.beir-v1.0.0-arguana.test.bge-base-en-v1.5.jsonl.gz \
+  -topicReader JsonStringVector \
+  -output runs/run.beir-v1.0.0-arguana.bge-base-en-v1.5.bge-flat-int8-cached.topics.beir-v1.0.0-arguana.test.bge-base-en-v1.5.jsonl.txt \
+  -hits 1000 -removeQuery -threads 16 &
+```
+
+Evaluation can be performed using `trec_eval`:
+
+```
+bin/trec_eval -c -m ndcg_cut.10 tools/topics-and-qrels/qrels.beir-v1.0.0-arguana.test.txt runs/run.beir-v1.0.0-arguana.bge-base-en-v1.5.bge-flat-int8-cached.topics.beir-v1.0.0-arguana.test.bge-base-en-v1.5.jsonl.txt
+bin/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.beir-v1.0.0-arguana.test.txt runs/run.beir-v1.0.0-arguana.bge-base-en-v1.5.bge-flat-int8-cached.topics.beir-v1.0.0-arguana.test.bge-base-en-v1.5.jsonl.txt
+bin/trec_eval -c -m recall.1000 tools/topics-and-qrels/qrels.beir-v1.0.0-arguana.test.txt runs/run.beir-v1.0.0-arguana.bge-base-en-v1.5.bge-flat-int8-cached.topics.beir-v1.0.0-arguana.test.bge-base-en-v1.5.jsonl.txt
+```
+
+## Effectiveness
+
+With the above commands, you should be able to reproduce the following results:
+
+| **nDCG@10**                                                                                                  | **BGE-base-en-v1.5**|
+|:-------------------------------------------------------------------------------------------------------------|-----------|
+| BEIR (v1.0.0): ArguAna                                                                                       | 0.6361    |
+| **R@100**                                                                                                    | **BGE-base-en-v1.5**|
+| BEIR (v1.0.0): ArguAna                                                                                       | 0.9915    |
+| **R@1000**                                                                                                   | **BGE-base-en-v1.5**|
+| BEIR (v1.0.0): ArguAna                                                                                       | 0.9964    |
+
+The above figures are from running brute-force search with cached queries on non-quantized flat indexes.
+With cached queries on quantized flat indexes, observed results may differ slightly (typically, lower), but scores should generally be within 0.004 of the results reported above (with some outliers).
+Note that quantization is non-deterministic due to sampling (i.e., results may differ slightly between trials).
diff --git a/...ions/regressions-beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.onnx.md b/...ions/regressions-beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.onnx.md
@@ -0,0 +1,84 @@
+# Anserini Regressions: BEIR (v1.0.0) &mdash; ArguAna
+
+**Model**: [BGE-base-en-v1.5](https://huggingface.co/BAAI/bge-base-en-v1.5) with quantized flat indexes (using ONNX for on-the-fly query encoding)
+
+This page describes regression experiments, integrated into Anserini's regression testing framework, using the [BGE-base-en-v1.5](https://huggingface.co/BAAI/bge-base-en-v1.5) model on [BEIR (v1.0.0) &mdash; ArguAna](http://beir.ai/), as described in the following paper:
+
+> Shitao Xiao, Zheng Liu, Peitian Zhang, and Niklas Muennighoff. [C-Pack: Packaged Resources To Advance General Chinese Embedding.](https://arxiv.org/abs/2309.07597) _arXiv:2309.07597_, 2023.
+
+In these experiments, we are using ONNX to perform query encoding on the fly.
+
+The exact configurations for these regressions are stored in [this YAML file](../../src/main/resources/regression/beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.onnx.yaml).
+Note that this page is automatically generated from [this template](../../src/main/resources/docgen/templates/beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.onnx.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 beir-v1.0.0-arguana.bge-base-en-v1.5.parquet.flat-int8.onnx
+```
+
+All the BEIR corpora, encoded by the BGE-base-en-v1.5 model and stored in Parquet format, are available for download:
+
+```bash
+wget https://rgw.cs.uwaterloo.ca/pyserini/data/beir-v1.0.0-bge-base-en-v1.5.parquet.tar -P collections/
+tar xvf collections/beir-v1.0.0-bge-base-en-v1.5.parquet.tar -C collections/
+```
+
+The tarball is 194 GB and has MD5 checksum `c279f9fc2464574b482ec53efcc1c487`.
+After download and unpacking the corpora, the `run_regression.py` command above should work without any issue.
+
+## Indexing
+
+Sample indexing command, building quantized flat indexes:
+
+```
+bin/run.sh io.anserini.index.IndexFlatDenseVectors \
+  -threads 16 \
+  -collection ParquetDenseVectorCollection \
+  -input /path/to/beir-v1.0.0-arguana.bge-base-en-v1.5 \
+  -generator ParquetDenseVectorDocumentGenerator \
+  -index indexes/lucene-flat-int8.beir-v1.0.0-arguana.bge-base-en-v1.5/ \
+  -quantize.int8 \
+  >& logs/log.beir-v1.0.0-arguana.bge-base-en-v1.5 &
+```
+
+The path `/path/to/beir-v1.0.0-arguana.bge-base-en-v1.5/` should point to the corpus downloaded above.
+
+## Retrieval
+
+Topics and qrels are stored [here](https://github.com/castorini/anserini-tools/tree/master/topics-and-qrels), 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.SearchFlatDenseVectors \
+  -index indexes/lucene-flat-int8.beir-v1.0.0-arguana.bge-base-en-v1.5/ \
+  -topics tools/topics-and-qrels/topics.beir-v1.0.0-arguana.test.tsv.gz \
+  -topicReader TsvString \
+  -output runs/run.beir-v1.0.0-arguana.bge-base-en-v1.5.bge-flat-int8-onnx.topics.beir-v1.0.0-arguana.test.txt \
+  -encoder BgeBaseEn15 -hits 1000 -removeQuery -threads 16 &
+```
+
+Evaluation can be performed using `trec_eval`:
+
+```
+bin/trec_eval -c -m ndcg_cut.10 tools/topics-and-qrels/qrels.beir-v1.0.0-arguana.test.txt runs/run.beir-v1.0.0-arguana.bge-base-en-v1.5.bge-flat-int8-onnx.topics.beir-v1.0.0-arguana.test.txt
+bin/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.beir-v1.0.0-arguana.test.txt runs/run.beir-v1.0.0-arguana.bge-base-en-v1.5.bge-flat-int8-onnx.topics.beir-v1.0.0-arguana.test.txt
+bin/trec_eval -c -m recall.1000 tools/topics-and-qrels/qrels.beir-v1.0.0-arguana.test.txt runs/run.beir-v1.0.0-arguana.bge-base-en-v1.5.bge-flat-int8-onnx.topics.beir-v1.0.0-arguana.test.txt
+```
+
+## Effectiveness
+
+With the above commands, you should be able to reproduce the following results:
+
+| **nDCG@10**                                                                                                  | **BGE-base-en-v1.5**|
+|:-------------------------------------------------------------------------------------------------------------|-----------|
+| BEIR (v1.0.0): ArguAna                                                                                       | 0.6361    |
+| **R@100**                                                                                                    | **BGE-base-en-v1.5**|
+| BEIR (v1.0.0): ArguAna                                                                                       | 0.9915    |
+| **R@1000**                                                                                                   | **BGE-base-en-v1.5**|
+| BEIR (v1.0.0): ArguAna                                                                                       | 0.9964    |
+
+The above figures are from running brute-force search with cached queries on non-quantized flat indexes.
+With ONNX query encoding on quantized flat indexes, observed results may differ slightly (typically, lower), but scores should generally be within 0.004 of the results reported above (with some outliers).
+Note that quantization is non-deterministic due to sampling (i.e., results may differ slightly between trials).