zeroshot's picture
Update README.md
03bae31
|
raw
history blame
12.6 kB
metadata
license: mit
language:
  - en
tags:
  - sparse
  - sparsity
  - quantized
  - onnx
  - embeddings
  - int8
  - mteb
  - deepsparse
model-index:
  - name: bge-large-en-v1.5-quant
    results:
      - task:
          type: Classification
        dataset:
          type: mteb/amazon_counterfactual
          name: MTEB AmazonCounterfactualClassification (en)
          config: en
          split: test
          revision: e8379541af4e31359cca9fbcf4b00f2671dba205
        metrics:
          - type: accuracy
            value: 75.53731343283583
          - type: ap
            value: 38.30609312253564
          - type: f1
            value: 69.42802757893695
      - task:
          type: STS
        dataset:
          type: mteb/biosses-sts
          name: MTEB BIOSSES
          config: default
          split: test
          revision: d3fb88f8f02e40887cd149695127462bbcf29b4a
        metrics:
          - type: cos_sim_pearson
            value: 89.27346145216443
          - type: cos_sim_spearman
            value: 88.36526647458979
          - type: euclidean_pearson
            value: 86.83053354694746
          - type: euclidean_spearman
            value: 87.56223612880584
          - type: manhattan_pearson
            value: 86.59250609226758
          - type: manhattan_spearman
            value: 87.70681773644885
      - task:
          type: STS
        dataset:
          type: mteb/sickr-sts
          name: MTEB SICK-R
          config: default
          split: test
          revision: a6ea5a8cab320b040a23452cc28066d9beae2cee
        metrics:
          - type: cos_sim_pearson
            value: 86.18998669716373
          - type: cos_sim_spearman
            value: 82.06129973984048
          - type: euclidean_pearson
            value: 83.65969509485801
          - type: euclidean_spearman
            value: 81.91666612708826
          - type: manhattan_pearson
            value: 83.6906794731384
          - type: manhattan_spearman
            value: 81.91752705367436
      - task:
          type: STS
        dataset:
          type: mteb/sts12-sts
          name: MTEB STS12
          config: default
          split: test
          revision: a0d554a64d88156834ff5ae9920b964011b16384
        metrics:
          - type: cos_sim_pearson
            value: 86.93407086985752
          - type: cos_sim_spearman
            value: 78.82992283957066
          - type: euclidean_pearson
            value: 83.39733473832982
          - type: euclidean_spearman
            value: 78.86999229850214
          - type: manhattan_pearson
            value: 83.39397058098533
          - type: manhattan_spearman
            value: 78.85397971200753
      - task:
          type: STS
        dataset:
          type: mteb/sts13-sts
          name: MTEB STS13
          config: default
          split: test
          revision: 7e90230a92c190f1bf69ae9002b8cea547a64cca
        metrics:
          - type: cos_sim_pearson
            value: 87.2586009863056
          - type: cos_sim_spearman
            value: 87.99415514558852
          - type: euclidean_pearson
            value: 86.98993652364359
          - type: euclidean_spearman
            value: 87.72725335668807
          - type: manhattan_pearson
            value: 86.897205761048
          - type: manhattan_spearman
            value: 87.65231103509018
      - task:
          type: STS
        dataset:
          type: mteb/sts14-sts
          name: MTEB STS14
          config: default
          split: test
          revision: 6031580fec1f6af667f0bd2da0a551cf4f0b2375
        metrics:
          - type: cos_sim_pearson
            value: 85.41417660460755
          - type: cos_sim_spearman
            value: 83.50291886604928
          - type: euclidean_pearson
            value: 84.67758839660924
          - type: euclidean_spearman
            value: 83.4368059512681
          - type: manhattan_pearson
            value: 84.66027228213025
          - type: manhattan_spearman
            value: 83.43472054456252
      - task:
          type: STS
        dataset:
          type: mteb/sts15-sts
          name: MTEB STS15
          config: default
          split: test
          revision: ae752c7c21bf194d8b67fd573edf7ae58183cbe3
        metrics:
          - type: cos_sim_pearson
            value: 88.02513262365703
          - type: cos_sim_spearman
            value: 89.00430907638267
          - type: euclidean_pearson
            value: 88.16290361497319
          - type: euclidean_spearman
            value: 88.6645154822661
          - type: manhattan_pearson
            value: 88.15337528825458
          - type: manhattan_spearman
            value: 88.66202950081507
      - task:
          type: STS
        dataset:
          type: mteb/sts16-sts
          name: MTEB STS16
          config: default
          split: test
          revision: 4d8694f8f0e0100860b497b999b3dbed754a0513
        metrics:
          - type: cos_sim_pearson
            value: 85.10194022827035
          - type: cos_sim_spearman
            value: 86.45367112223394
          - type: euclidean_pearson
            value: 85.45292931769094
          - type: euclidean_spearman
            value: 86.06607589083283
          - type: manhattan_pearson
            value: 85.4111233047049
          - type: manhattan_spearman
            value: 86.04379654118996
      - task:
          type: STS
        dataset:
          type: mteb/sts17-crosslingual-sts
          name: MTEB STS17 (en-en)
          config: en-en
          split: test
          revision: af5e6fb845001ecf41f4c1e033ce921939a2a68d
        metrics:
          - type: cos_sim_pearson
            value: 89.86966589113663
          - type: cos_sim_spearman
            value: 89.5617056243649
          - type: euclidean_pearson
            value: 89.018495917952
          - type: euclidean_spearman
            value: 88.387335721179
          - type: manhattan_pearson
            value: 89.07568042943448
          - type: manhattan_spearman
            value: 88.51733863475219
      - task:
          type: STS
        dataset:
          type: mteb/sts22-crosslingual-sts
          name: MTEB STS22 (en)
          config: en
          split: test
          revision: 6d1ba47164174a496b7fa5d3569dae26a6813b80
        metrics:
          - type: cos_sim_pearson
            value: 68.38465344518238
          - type: cos_sim_spearman
            value: 68.15219488291783
          - type: euclidean_pearson
            value: 68.99169681132668
          - type: euclidean_spearman
            value: 68.01334641045888
          - type: manhattan_pearson
            value: 68.84952679202642
          - type: manhattan_spearman
            value: 67.85430179655137
      - task:
          type: STS
        dataset:
          type: mteb/stsbenchmark-sts
          name: MTEB STSBenchmark
          config: default
          split: test
          revision: b0fddb56ed78048fa8b90373c8a3cfc37b684831
        metrics:
          - type: cos_sim_pearson
            value: 86.60574360222778
          - type: cos_sim_spearman
            value: 87.8878986593873
          - type: euclidean_pearson
            value: 87.11557232168404
          - type: euclidean_spearman
            value: 87.40944677043365
          - type: manhattan_pearson
            value: 87.10395398212532
          - type: manhattan_spearman
            value: 87.35977283466168
      - task:
          type: PairClassification
        dataset:
          type: mteb/sprintduplicatequestions-pairclassification
          name: MTEB SprintDuplicateQuestions
          config: default
          split: test
          revision: d66bd1f72af766a5cc4b0ca5e00c162f89e8cc46
        metrics:
          - type: cos_sim_accuracy
            value: 99.84752475247525
          - type: cos_sim_ap
            value: 96.49316696572335
          - type: cos_sim_f1
            value: 92.35352532274081
          - type: cos_sim_precision
            value: 91.71597633136095
          - type: cos_sim_recall
            value: 93
          - type: dot_accuracy
            value: 99.77326732673268
          - type: dot_ap
            value: 93.5497681978726
          - type: dot_f1
            value: 88.35582208895552
          - type: dot_precision
            value: 88.31168831168831
          - type: dot_recall
            value: 88.4
          - type: euclidean_accuracy
            value: 99.84653465346534
          - type: euclidean_ap
            value: 96.36378999360083
          - type: euclidean_f1
            value: 92.33052944087086
          - type: euclidean_precision
            value: 91.38099902056807
          - type: euclidean_recall
            value: 93.30000000000001
          - type: manhattan_accuracy
            value: 99.84455445544555
          - type: manhattan_ap
            value: 96.36035171233175
          - type: manhattan_f1
            value: 92.13260761999011
          - type: manhattan_precision
            value: 91.1851126346719
          - type: manhattan_recall
            value: 93.10000000000001
          - type: max_accuracy
            value: 99.84752475247525
          - type: max_ap
            value: 96.49316696572335
          - type: max_f1
            value: 92.35352532274081
      - task:
          type: PairClassification
        dataset:
          type: mteb/twittersemeval2015-pairclassification
          name: MTEB TwitterSemEval2015
          config: default
          split: test
          revision: 70970daeab8776df92f5ea462b6173c0b46fd2d1
        metrics:
          - type: cos_sim_accuracy
            value: 87.26828396018358
          - type: cos_sim_ap
            value: 77.79878217023162
          - type: cos_sim_f1
            value: 71.0425694621463
          - type: cos_sim_precision
            value: 68.71301775147928
          - type: cos_sim_recall
            value: 73.53562005277044
          - type: dot_accuracy
            value: 84.01978899684092
          - type: dot_ap
            value: 66.12134149171163
          - type: dot_f1
            value: 63.283507097098365
          - type: dot_precision
            value: 60.393191081275475
          - type: dot_recall
            value: 66.46437994722955
          - type: euclidean_accuracy
            value: 87.24444179531503
          - type: euclidean_ap
            value: 77.84821131946212
          - type: euclidean_f1
            value: 71.30456661215247
          - type: euclidean_precision
            value: 68.1413801394566
          - type: euclidean_recall
            value: 74.77572559366754
          - type: manhattan_accuracy
            value: 87.19079692436074
          - type: manhattan_ap
            value: 77.78054941055291
          - type: manhattan_f1
            value: 71.13002127393318
          - type: manhattan_precision
            value: 67.65055939062128
          - type: manhattan_recall
            value: 74.9868073878628
          - type: max_accuracy
            value: 87.26828396018358
          - type: max_ap
            value: 77.84821131946212
          - type: max_f1
            value: 71.30456661215247
      - task:
          type: PairClassification
        dataset:
          type: mteb/twitterurlcorpus-pairclassification
          name: MTEB TwitterURLCorpus
          config: default
          split: test
          revision: 8b6510b0b1fa4e4c4f879467980e9be563ec1cdf
        metrics:
          - type: cos_sim_accuracy
            value: 88.91023402025847
          - type: cos_sim_ap
            value: 85.94088151184411
          - type: cos_sim_f1
            value: 78.25673997223645
          - type: cos_sim_precision
            value: 74.45433059919367
          - type: cos_sim_recall
            value: 82.46843239913767
          - type: dot_accuracy
            value: 87.91865564481701
          - type: dot_ap
            value: 82.75373957440969
          - type: dot_f1
            value: 75.97383507276201
          - type: dot_precision
            value: 72.67294713160854
          - type: dot_recall
            value: 79.5888512473052
          - type: euclidean_accuracy
            value: 88.8539604921023
          - type: euclidean_ap
            value: 85.71590936389937
          - type: euclidean_f1
            value: 77.82902261742242
          - type: euclidean_precision
            value: 74.7219270279844
          - type: euclidean_recall
            value: 81.20572836464429
          - type: manhattan_accuracy
            value: 88.78992509799356
          - type: manhattan_ap
            value: 85.70200619366904
          - type: manhattan_f1
            value: 77.85875848203065
          - type: manhattan_precision
            value: 72.94315506222671
          - type: manhattan_recall
            value: 83.48475515860795
          - type: max_accuracy
            value: 88.91023402025847
          - type: max_ap
            value: 85.94088151184411
          - type: max_f1
            value: 78.25673997223645

bge-large-en-v1.5-quant

latency

DeepSparse is able to improve latency performance on a 10 core laptop by 3X and up to 5X on a 16 core AWS instance.

Usage

This is the quantized (INT8) ONNX variant of the bge-large-en-v1.5 embeddings model accelerated with Sparsify for quantization and DeepSparseSentenceTransformers for inference.

pip install -U deepsparse-nightly[sentence_transformers]
from deepsparse.sentence_transformers import DeepSparseSentenceTransformer
model = DeepSparseSentenceTransformer('zeroshot/bge-large-en-v1.5-quant', export=False)

# Our sentences we like to encode
sentences = ['This framework generates embeddings for each input sentence',
    'Sentences are passed as a list of string.',
    'The quick brown fox jumps over the lazy dog.']

# Sentences are encoded by calling model.encode()
embeddings = model.encode(sentences)

# Print the embeddings
for sentence, embedding in zip(sentences, embeddings):
    print("Sentence:", sentence)
    print("Embedding:", embedding.shape)
    print("")

For general questions on these models and sparsification methods, reach out to the engineering team on our community Slack.