Upload folder using huggingface_hub

lucene-hnsw/bge-base-en-v1.5/lucene-hnsw-int8.msmarco-v1-passage.bge-base-en-v1.5.20240117.53514b.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02e9d095f82a325f4ff341cf3cc72c6366557716e2f6c59faa0cc8e84513b53f
+size 31118393425

lucene-hnsw/bge-base-en-v1.5/lucene-hnsw.msmarco-v1-passage.bge-base-en-v1.5.20240117.53514b.README.md

@@ -0,0 +1,28 @@
+# msmarco-v1-passage.bge-base-en-v1.5
+
+Lucene HNSW indexes of MS MARCO V1 Passage using BGE-base-en-v1.5.
+
+These indexes were built 2024/01/17 on `orca` at Anserini commit [`53514b`](https://github.com/castorini/anserini/commit/53514b1ab29398a4bb6ff4a315b7394e509e6be5) (2024/01/13), with Lucene 9.9.1.
+
+Here are the indexing commands for the non-quantized and quantized versions:
+
+```bash
+target/appassembler/bin/IndexHnswDenseVectors \
+  -collection JsonDenseVectorCollection \
+  -input /mnt/collections/msmarco/msmarco-passage-bge-base-en-v1.5/ \
+  -generator HnswDenseVectorDocumentGenerator \
+  -index indexes/lucene-hnsw.msmarco-passage-bge-base-en-v1.5.efC1000.1/ \
+  -threads 8 -M 16 -efC 1000 -memoryBuffer 65536 -optimize \
+  >& logs/log.msmarco-passage-bge-base-en-v1.5.efC1000.1 &
+
+target/appassembler/bin/IndexHnswDenseVectors \
+  -collection JsonDenseVectorCollection \
+  -input /mnt/collections/msmarco/msmarco-passage-bge-base-en-v1.5/ \
+  -generator HnswDenseVectorDocumentGenerator \
+  -index indexes/lucene-hnsw.msmarco-passage-bge-base-en-v1.5-int8.efC1000.1/ \
+  -threads 8 -M 16 -efC 1000 -memoryBuffer 65536 -optimize -quantize.int8 \
+  >& logs/log.msmarco-passage-bge-base-en-v1.5-int8.efC1000.1 &
+```
+
+I ran four trials and picked the index instance that yielded the highest retrieval scores.
+Most of the trials yielded scores that were close; I selected the "best" based on eyeballing.

lucene-hnsw/bge-base-en-v1.5/lucene-hnsw.msmarco-v1-passage.bge-base-en-v1.5.20240117.53514b.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1daf449b870543887b711d8280f4a1aa4e268bed85ac37f130e94895901a261b
+size 25993352262

lucene-hnsw/cohere-embed-english/lucene-hnsw-int8.msmarco-v1-passage.cohere-embed-english-v3.0.20240228.eacd13.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a9cca6511890b05bdb21b97c26dd46796e6bb48e8ea0de9d4110f8111e2c1bd
+size 29167754602

lucene-hnsw/cohere-embed-english/lucene-hnsw.msmarco-v1-passage.cohere-embed-english-v3.0.20240228.eacd13.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad4f2df730dbba8dc2a4f30ceaccbe01418b0743795c97dc380ca38d8baeab9c
+size 22094914552

lucene-hnsw/cosdpr-distil/lucene-hnsw-int8.msmarco-v1-passage.cosdpr-distil.20240108.825148.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fd79f24980e0a72fedd0887f1bedcdba46e90253b81bef366ef2dcb57ac733d
+size 30140494326

lucene-hnsw/cosdpr-distil/lucene-hnsw.msmarco-v1-passage.cosdpr-distil.20240108.825148.README.md

@@ -0,0 +1,28 @@
+# msmarco-v1-passage.cosdpr-distil
+
+Lucene HNSW indexes of MS MARCO V1 Passage using cosDPR-distil.
+
+These indexes were built 2024/01/07 on `orca` at Anserini commit [`825148`](https://github.com/castorini/anserini/commit/825148afba0303276c37dd838be897b8443d9774) (2023/12/24), with Lucene 9.9.1.
+
+Here are the indexing commands for the non-quantized and quantized versions:
+
+```bash
+target/appassembler/bin/IndexHnswDenseVectors \
+  -collection JsonDenseVectorCollection \
+  -input /mnt/collections/msmarco/msmarco-passage-cos-dpr-distil/ \
+  -generator HnswDenseVectorDocumentGenerator \
+  -index indexes/lucene-hnsw.msmarco-passage-cos-dpr-distil.efC1000.1/ \
+  -threads 8 -M 16 -efC 1000 -memoryBuffer 65536 -optimize \
+  >& logs/log.msmarco-passage-cos-dpr-distil.efC1000.1 &
+
+target/appassembler/bin/IndexHnswDenseVectors \
+  -collection JsonDenseVectorCollection \
+  -input /mnt/collections/msmarco/msmarco-passage-cos-dpr-distil/ \
+  -generator HnswDenseVectorDocumentGenerator \
+  -index indexes/lucene-hnsw.msmarco-passage-cos-dpr-distil-int8.efC1000.1/ \
+  -threads 8 -M 16 -efC 1000 -memoryBuffer 65536 -optimize -quantize.int8 \
+  >& logs/log.msmarco-passage-cos-dpr-distil-int8.efC1000.1 &
+```
+
+I ran four trials and picked the index instance that yielded the highest retrieval scores.
+Most of the trials yielded scores that were close; I selected the "best" based on eyeballing.

lucene-hnsw/cosdpr-distil/lucene-hnsw.msmarco-v1-passage.cosdpr-distil.20240108.825148.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1147b02411452621455ce97d28011be84f38a11e502f371654ed03c084e1085d
+size 24632173689