CoolFace
Modelpublic

zeroshot/gte-large-sparse

sourceHugging Facemitupdated 3y agoView on Hugging Face
2likes35downloads
README.md430 linesDownload Raw Back to root
1---2tags:3- sparse sparsity quantized onnx embeddings int84- mteb5model-index:6- name: gte-large-sparse7  results:8  - task:9      type: STS10    dataset:11      type: mteb/biosses-sts12      name: MTEB BIOSSES13      config: default14      split: test15      revision: d3fb88f8f02e40887cd149695127462bbcf29b4a16    metrics:17    - type: cos_sim_pearson18      value: 88.6425341092821419    - type: cos_sim_spearman20      value: 85.8338834941065221    - type: euclidean_pearson22      value: 86.8612615931873523    - type: euclidean_spearman24      value: 85.6158062359116325    - type: manhattan_pearson26      value: 86.690113288338327    - type: manhattan_spearman28      value: 85.6025529218776929  - task:30      type: STS31    dataset:32      type: mteb/sickr-sts33      name: MTEB SICK-R34      config: default35      split: test36      revision: a6ea5a8cab320b040a23452cc28066d9beae2cee37    metrics:38    - type: cos_sim_pearson39      value: 85.2331464059160740    - type: cos_sim_spearman41      value: 79.0007854510433842    - type: euclidean_pearson43      value: 83.4800925450071444    - type: euclidean_spearman45      value: 78.9541300138993946    - type: manhattan_pearson47      value: 83.4694556602594148    - type: manhattan_spearman49      value: 78.924170720813550  - task:51      type: STS52    dataset:53      type: mteb/sts12-sts54      name: MTEB STS1255      config: default56      split: test57      revision: a0d554a64d88156834ff5ae9920b964011b1638458    metrics:59    - type: cos_sim_pearson60      value: 81.7752666604380461    - type: cos_sim_spearman62      value: 73.484906328586763    - type: euclidean_pearson64      value: 78.0447793274052465    - type: euclidean_spearman66      value: 73.0139420577174367    - type: manhattan_pearson68      value: 78.0883668450329469    - type: manhattan_spearman70      value: 73.0507471109814971  - task:72      type: STS73    dataset:74      type: mteb/sts13-sts75      name: MTEB STS1376      config: default77      split: test78      revision: 7e90230a92c190f1bf69ae9002b8cea547a64cca79    metrics:80    - type: cos_sim_pearson81      value: 84.5783921566135282    - type: cos_sim_spearman83      value: 86.1385476734515384    - type: euclidean_pearson85      value: 85.1271260994644986    - type: euclidean_spearman87      value: 85.5249799478902688    - type: manhattan_pearson89      value: 85.0683314161117390    - type: manhattan_spearman91      value: 85.4500306863646692  - task:93      type: STS94    dataset:95      type: mteb/sts14-sts96      name: MTEB STS1497      config: default98      split: test99      revision: 6031580fec1f6af667f0bd2da0a551cf4f0b2375100    metrics:101    - type: cos_sim_pearson102      value: 83.30485126978374103    - type: cos_sim_spearman104      value: 80.36497172462357105    - type: euclidean_pearson106      value: 82.91977909424605107    - type: euclidean_spearman108      value: 80.16995106297438109    - type: manhattan_pearson110      value: 82.88200991402184111    - type: manhattan_spearman112      value: 80.14259757215227113  - task:114      type: STS115    dataset:116      type: mteb/sts15-sts117      name: MTEB STS15118      config: default119      split: test120      revision: ae752c7c21bf194d8b67fd573edf7ae58183cbe3121    metrics:122    - type: cos_sim_pearson123      value: 86.99883111314007124    - type: cos_sim_spearman125      value: 88.531352572377126    - type: euclidean_pearson127      value: 87.96834578059067128    - type: euclidean_spearman129      value: 88.44800718542935130    - type: manhattan_pearson131      value: 87.94889391725033132    - type: manhattan_spearman133      value: 88.45467695837115134  - task:135      type: STS136    dataset:137      type: mteb/sts16-sts138      name: MTEB STS16139      config: default140      split: test141      revision: 4d8694f8f0e0100860b497b999b3dbed754a0513142    metrics:143    - type: cos_sim_pearson144      value: 82.4636984892402145    - type: cos_sim_spearman146      value: 84.0808920789148147    - type: euclidean_pearson148      value: 83.70613486028309149    - type: euclidean_spearman150      value: 84.35941626905009151    - type: manhattan_pearson152      value: 83.70259457073782153    - type: manhattan_spearman154      value: 84.35496521501604155  - task:156      type: STS157    dataset:158      type: mteb/sts17-crosslingual-sts159      name: MTEB STS17 (en-en)160      config: en-en161      split: test162      revision: af5e6fb845001ecf41f4c1e033ce921939a2a68d163    metrics:164    - type: cos_sim_pearson165      value: 88.76172944971023166    - type: cos_sim_spearman167      value: 89.4190945039165168    - type: euclidean_pearson169      value: 89.47263005347381170    - type: euclidean_spearman171      value: 89.49228360724095172    - type: manhattan_pearson173      value: 89.49959868816694174    - type: manhattan_spearman175      value: 89.5314536157954176  - task:177      type: STS178    dataset:179      type: mteb/sts22-crosslingual-sts180      name: MTEB STS22 (en)181      config: en182      split: test183      revision: 6d1ba47164174a496b7fa5d3569dae26a6813b80184    metrics:185    - type: cos_sim_pearson186      value: 64.57158223787549187    - type: cos_sim_spearman188      value: 66.75053533168037189    - type: euclidean_pearson190      value: 66.45526604831747191    - type: euclidean_spearman192      value: 66.14567667353113193    - type: manhattan_pearson194      value: 66.47352000151176195    - type: manhattan_spearman196      value: 66.21099856852885197  - task:198      type: STS199    dataset:200      type: mteb/stsbenchmark-sts201      name: MTEB STSBenchmark202      config: default203      split: test204      revision: b0fddb56ed78048fa8b90373c8a3cfc37b684831205    metrics:206    - type: cos_sim_pearson207      value: 85.055653571006208    - type: cos_sim_spearman209      value: 85.45387832634702210    - type: euclidean_pearson211      value: 86.31667154906651212    - type: euclidean_spearman213      value: 85.66079590537946214    - type: manhattan_pearson215      value: 86.2806853257308216    - type: manhattan_spearman217      value: 85.63700636713952218  - task:219      type: PairClassification220    dataset:221      type: mteb/sprintduplicatequestions-pairclassification222      name: MTEB SprintDuplicateQuestions223      config: default224      split: test225      revision: d66bd1f72af766a5cc4b0ca5e00c162f89e8cc46226    metrics:227    - type: cos_sim_accuracy228      value: 99.78811881188119229    - type: cos_sim_ap230      value: 94.67027715905307231    - type: cos_sim_f1232      value: 89.33074684772066233    - type: cos_sim_precision234      value: 86.7231638418079235    - type: cos_sim_recall236      value: 92.10000000000001237    - type: dot_accuracy238      value: 99.47128712871287239    - type: dot_ap240      value: 78.41478815918727241    - type: dot_f1242      value: 73.30049261083744243    - type: dot_precision244      value: 72.23300970873787245    - type: dot_recall246      value: 74.4247    - type: euclidean_accuracy248      value: 99.78415841584159249    - type: euclidean_ap250      value: 94.60075930867181251    - type: euclidean_f1252      value: 89.12175648702593253    - type: euclidean_precision254      value: 88.94422310756973255    - type: euclidean_recall256      value: 89.3257    - type: manhattan_accuracy258      value: 99.78415841584159259    - type: manhattan_ap260      value: 94.62867439278095261    - type: manhattan_f1262      value: 89.2337536372454263    - type: manhattan_precision264      value: 86.62900188323917265    - type: manhattan_recall266      value: 92.0267    - type: max_accuracy268      value: 99.78811881188119269    - type: max_ap270      value: 94.67027715905307271    - type: max_f1272      value: 89.33074684772066273  - task:274      type: PairClassification275    dataset:276      type: mteb/twittersemeval2015-pairclassification277      name: MTEB TwitterSemEval2015278      config: default279      split: test280      revision: 70970daeab8776df92f5ea462b6173c0b46fd2d1281    metrics:282    - type: cos_sim_accuracy283      value: 85.09864695714371284    - type: cos_sim_ap285      value: 70.33704198164713286    - type: cos_sim_f1287      value: 66.22893954410307288    - type: cos_sim_precision289      value: 62.42410088743577290    - type: cos_sim_recall291      value: 70.52770448548813292    - type: dot_accuracy293      value: 79.11426357513263294    - type: dot_ap295      value: 49.15484584572233296    - type: dot_f1297      value: 51.12580243364951298    - type: dot_precision299      value: 40.13840830449827300    - type: dot_recall301      value: 70.3957783641161302    - type: euclidean_accuracy303      value: 85.15825236931514304    - type: euclidean_ap305      value: 70.51017350854076306    - type: euclidean_f1307      value: 66.45416294785159308    - type: euclidean_precision309      value: 64.29805082654823310    - type: euclidean_recall311      value: 68.7598944591029312    - type: manhattan_accuracy313      value: 85.1403707456637314    - type: manhattan_ap315      value: 70.47587863399994316    - type: manhattan_f1317      value: 66.4576802507837318    - type: manhattan_precision319      value: 63.32138590203107320    - type: manhattan_recall321      value: 69.92084432717678322    - type: max_accuracy323      value: 85.15825236931514324    - type: max_ap325      value: 70.51017350854076326    - type: max_f1327      value: 66.4576802507837328  - task:329      type: PairClassification330    dataset:331      type: mteb/twitterurlcorpus-pairclassification332      name: MTEB TwitterURLCorpus333      config: default334      split: test335      revision: 8b6510b0b1fa4e4c4f879467980e9be563ec1cdf336    metrics:337    - type: cos_sim_accuracy338      value: 88.8539604921023339    - type: cos_sim_ap340      value: 85.71869912577101341    - type: cos_sim_f1342      value: 78.00535626720983343    - type: cos_sim_precision344      value: 76.46232344893885345    - type: cos_sim_recall346      value: 79.61194949183862347    - type: dot_accuracy348      value: 84.57717235223348349    - type: dot_ap350      value: 74.89496650237145351    - type: dot_f1352      value: 69.05327823892932353    - type: dot_precision354      value: 65.75666829166377355    - type: dot_recall356      value: 72.69787496150293357    - type: euclidean_accuracy358      value: 88.89471028835332359    - type: euclidean_ap360      value: 85.75169460500409361    - type: euclidean_f1362      value: 78.17055393586006363    - type: euclidean_precision364      value: 74.21118184334348365    - type: euclidean_recall366      value: 82.57622420696026367    - type: manhattan_accuracy368      value: 88.92187681918733369    - type: manhattan_ap370      value: 85.7496679471825371    - type: manhattan_f1372      value: 78.11088295687884373    - type: manhattan_precision374      value: 75.82083061535117375    - type: manhattan_recall376      value: 80.5435786880197377    - type: max_accuracy378      value: 88.92187681918733379    - type: max_ap380      value: 85.75169460500409381    - type: max_f1382      value: 78.17055393586006383license: mit384language:385- en386---387 388# gte-large-sparse389 390This is the sparse ONNX variant of the [gte-large](https://huggingface.co/thenlper/gte-large) embeddings model created with [DeepSparse Optimum](https://github.com/neuralmagic/optimum-deepsparse) for ONNX export/inference and Neural Magic's [Sparsify](https://github.com/neuralmagic/sparsify) for one-shot quantization (INT8) and unstructured pruning 50%.391 392Current list of sparse and quantized gte ONNX models:393 394| Links                                                                                               | Sparsification Method |395| --------------------------------------------------------------------------------------------------- | ---------------------- |396| [zeroshot/gte-large-sparse](https://huggingface.co/zeroshot/gte-large-sparse)     |    Quantization (INT8) & 50% Pruning                    |397| [zeroshot/gte-large-quant](https://huggingface.co/zeroshot/gte-large-quant)     |   Quantization (INT8)                     |398| [zeroshot/gte-base-sparse](https://huggingface.co/zeroshot/gte-base-sparse)     |    Quantization (INT8) & 50% Pruning                    |399| [zeroshot/gte-base-quant](https://huggingface.co/zeroshot/gte-base-quant)     |   Quantization (INT8)                     |400| [zeroshot/gte-small-sparse](https://huggingface.co/zeroshot/gte-small-sparse)     |    Quantization (INT8) & 50% Pruning                    |401| [zeroshot/gte-small-quant](https://huggingface.co/zeroshot/gte-small-quant)     |   Quantization (INT8)                     |402 403```bash404pip install -U deepsparse-nightly[sentence_transformers]405```406 407```python408from deepsparse.sentence_transformers import SentenceTransformer409model = SentenceTransformer('zeroshot/gte-large-sparse', export=False)410 411# Our sentences we like to encode412sentences = ['This framework generates embeddings for each input sentence',413    'Sentences are passed as a list of string.',414    'The quick brown fox jumps over the lazy dog.']415 416# Sentences are encoded by calling model.encode()417embeddings = model.encode(sentences)418 419# Print the embeddings420for sentence, embedding in zip(sentences, embeddings):421    print("Sentence:", sentence)422    print("Embedding:", embedding.shape)423    print("")424```425 426For further details regarding DeepSparse & Sentence Transformers integration, refer to the [DeepSparse README](https://github.com/neuralmagic/deepsparse/tree/main/src/deepsparse/sentence_transformers).427 428For general questions on these models and sparsification methods, reach out to the engineering team on our [community Slack](https://join.slack.com/t/discuss-neuralmagic/shared_invite/zt-q1a1cnvo-YBoICSIw3L1dmQpjBeDurQ).429 430![;)](https://media.giphy.com/media/bYg33GbNbNIVzSrr84/giphy-downsized-large.gif)