Reproduction log msmarco combined all [Anserini]

[ShanaxWorld]

#start-here-repro-log
System :
OS: Ubuntu (WSL2 on Windows)
Python: 3.11 (conda environment)
Java: OpenJDK 21
Hardware: Intel Core i5 9400F x64 architecture CPU (6 core CPU)
Ram : 16 GB
Graphics card : Nvidia GTX 1650 Super (4 GB)

Results:
1000000 58716381 374524070 collections/msmarco-passage/collection_jsonl/docs00.json
1000000 59072018 377845773 collections/msmarco-passage/collection_jsonl/docs01.json
1000000 58895092 375856044 collections/msmarco-passage/collection_jsonl/docs02.json
1000000 59277129 377452947 collections/msmarco-passage/collection_jsonl/docs03.json
1000000 59408028 378277584 collections/msmarco-passage/collection_jsonl/docs04.json
1000000 60659246 383758389 collections/msmarco-passage/collection_jsonl/docs05.json
1000000 63196730 400184520 collections/msmarco-passage/collection_jsonl/docs06.json
1000000 56920456 364726419 collections/msmarco-passage/collection_jsonl/docs07.json
841823 47767342 306155721 collections/msmarco-passage/collection_jsonl/docs08.json
8841823 523912422 3338781467 total
7437 29748 143300 collections/msmarco-passage/qrels.dev.small.tsv

Notes :
Had to setup ubuntu to allow WSL to work on the computer
Downloaded the collection from msmarco website
Install python and java sdk
Results reproduced without much of a problem

#3174
Reproduced BM25 baseline for MS MARCO passage ranking (Anserini sparse retrieval).

Environment:

Windows 11 (WSL2 Ubuntu)
Java 21
Python 3.12.3
Results:

MRR@10 = 0.1874 (dev.small)
Notes:

Installed Ubuntu (WSL2) to run the pipeline on Windows
Configured Python so that python maps to Python 3
Encountered CRLF issues when executing scripts from /mnt filesystem (fixed using dos2unix)
Observed file permission warnings (utime) on /mnt/e, but they did not affect processing
Maven build initially failed due to memory limits; resolved by skipping tests
Java 21 required for successful compilation
Successfully built Anserini index and retrieved ~7M results
Everything worked as expected after resolving the above issues.

#3175
Successfully reproduced BM25 baseline on the MS MARCO passage ranking task using the prebuilt Anserini index.

Environment
OS: Windows 11 (WSL2 Ubuntu)
Java: 21
Python: 3.12.3

Results
MRR@10 = 0.1875 (dev.small)

Notes
Initial runs with default settings (-parallelism 4, -hits 1000) resulted in the process being killed due to memory limits under WSL2.
Resolved by:
Reducing parallelism to 1
Reducing hits to 100
Setting Java heap: export JAVA_OPTS="-Xms512m -Xmx2g"
After these adjustments, BM25 retrieval completed successfully.

Dense Retrieval (BGE + HNSW)

Successfully reproduced dense retrieval using prebuilt HNSW index with BGE-base embeddings.

Results
MRR@10 = 0.3521 (dev.small)

Notes
Downloading the prebuilt dense index (~2–3GB) took approximately 30 minutes, depending on network conditions.
Query processing took ~37 minutes (~3 queries/sec), significantly slower than BM25.
Dense retrieval did not encounter memory issues under the same environment.
Results are consistent with expected performance for BGE-base on MS MARCO.

[lintool]

@ShanaxWorld please resolve conflicts. also missing a repro entry.

[ShanaxWorld]

@ShanaxWorld please resolve conflicts. also missing a repro entry.

Fixed it