regression.dl23-passage.splade-pp-sd.onnx.yaml Maven / Gradle / Ivy

Go to download
---
corpus: msmarco-v2-passage-splade-pp-sd
corpus_path: collections/msmarco/msmarco_v2_passage_splade_pp_sd/

download_url: https://rgw.cs.uwaterloo.ca/pyserini/data/msmarco_v2_passage_splade_pp_sd.tar
download_checksum: 061930dd615c7c807323ea7fc7957877
download_corpus: msmarco_v2_passage_splade_pp_sd

index_path: indexes/lucene-inverted.msmarco-v2-passage.splade-pp-sd/
collection_class: JsonVectorCollection
generator_class: DefaultLuceneDocumentGenerator
index_threads: 24
index_options: -impact -pretokenized -storeRaw
index_stats:
  documents: 138364198
  documents (non-empty): 138364198
  total terms: 907527565158

metrics:
  - metric: MAP@100
    command: bin/trec_eval
    params: -c -M 100 -m map -l 2
    separator: "\t"
    parse_index: 2
    metric_precision: 4
    can_combine: false
  - metric: MRR@100
    command: bin/trec_eval
    params: -c -M 100 -m recip_rank -l 2
    separator: "\t"
    parse_index: 2
    metric_precision: 4
    can_combine: false
  - metric: nDCG@10
    command: bin/trec_eval
    params: -c -m ndcg_cut.10
    separator: "\t"
    parse_index: 2
    metric_precision: 4
    can_combine: false
  - metric: R@100
    command: bin/trec_eval
    params: -c -m recall.100 -l 2
    separator: "\t"
    parse_index: 2
    metric_precision: 4
    can_combine: false
  - metric: R@1000
    command: bin/trec_eval
    params: -c -m recall.1000 -l 2
    separator: "\t"
    parse_index: 2
    metric_precision: 4
    can_combine: false

topic_reader: TsvInt
topics:
  - name: "[DL23 (Passage)](https://microsoft.github.io/msmarco/TREC-Deep-Learning)"
    id: dl23
    path: topics.dl23.txt
    qrel: qrels.dl23-passage.txt

models:
  - name: splade-pp-sd-onnx
    display: SPLADE++ CoCondenser-SelfDistil
    params: -parallelism 16 -impact -pretokenized -encoder SpladePlusPlusSelfDistil
    results:
      MAP@100:
        - 0.1963
      MRR@100:
        - 0.6985
      nDCG@10:
        - 0.4768
      R@100:
        - 0.4137
      R@1000:
        - 0.6731
  - name: splade-pp-sd-onnx+rm3
    display: +RM3
    params: -parallelism 16 -impact -pretokenized -encoder SpladePlusPlusSelfDistil -rm3 -collection JsonVectorCollection
    results:
      MAP@100:
        - 0.1938
      MRR@100:
        - 0.6273
      nDCG@10:
        - 0.4487
      R@100:
        - 0.3892
      R@1000:
        - 0.6486
  - name: splade-pp-sd-onnx+rocchio
    display: +Rocchio
    params: -parallelism 16 -impact -pretokenized -encoder SpladePlusPlusSelfDistil -rocchio -collection JsonVectorCollection
    results:
      MAP@100:
        - 0.2065
      MRR@100:
        - 0.7100
      nDCG@10:
        - 0.4774
      R@100:
        - 0.4226
      R@1000:
        - 0.7056