Model: 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 model on BEIR (v1.0.0) — CQADupStack-tex, as described in the following paper:
Shitao Xiao, Zheng Liu, Peitian Zhang, and Niklas Muennighoff. C-Pack: Packaged Resources To Advance General Chinese Embedding. 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.
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 beir-v1.0.0-cqadupstack-tex.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:
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.
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-cqadupstack-tex.bge-base-en-v1.5 \
-generator ParquetDenseVectorDocumentGenerator \
-index indexes/lucene-flat-int8.beir-v1.0.0-cqadupstack-tex.bge-base-en-v1.5/ \
-quantize.int8 \
>& logs/log.beir-v1.0.0-cqadupstack-tex.bge-base-en-v1.5 &
The path /path/to/beir-v1.0.0-cqadupstack-tex.bge-base-en-v1.5/ should point to the corpus downloaded above.
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.SearchFlatDenseVectors \
-index indexes/lucene-flat-int8.beir-v1.0.0-cqadupstack-tex.bge-base-en-v1.5/ \
-topics tools/topics-and-qrels/topics.beir-v1.0.0-cqadupstack-tex.test.tsv.gz \
-topicReader TsvString \
-output runs/run.beir-v1.0.0-cqadupstack-tex.bge-base-en-v1.5.bge-flat-int8-onnx.topics.beir-v1.0.0-cqadupstack-tex.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-cqadupstack-tex.test.txt runs/run.beir-v1.0.0-cqadupstack-tex.bge-base-en-v1.5.bge-flat-int8-onnx.topics.beir-v1.0.0-cqadupstack-tex.test.txt
bin/trec_eval -c -m recall.100 tools/topics-and-qrels/qrels.beir-v1.0.0-cqadupstack-tex.test.txt runs/run.beir-v1.0.0-cqadupstack-tex.bge-base-en-v1.5.bge-flat-int8-onnx.topics.beir-v1.0.0-cqadupstack-tex.test.txt
bin/trec_eval -c -m recall.1000 tools/topics-and-qrels/qrels.beir-v1.0.0-cqadupstack-tex.test.txt runs/run.beir-v1.0.0-cqadupstack-tex.bge-base-en-v1.5.bge-flat-int8-onnx.topics.beir-v1.0.0-cqadupstack-tex.test.txt
With the above commands, you should be able to reproduce the following results:
| nDCG@10 | BGE-base-en-v1.5 |
|---|---|
| BEIR (v1.0.0): CQADupStack-tex | 0.3115 |
| R@100 | BGE-base-en-v1.5 |
| BEIR (v1.0.0): CQADupStack-tex | 0.6486 |
| R@1000 | BGE-base-en-v1.5 |
| BEIR (v1.0.0): CQADupStack-tex | 0.8537 |
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).