SGPT-2.7B-weightedmean-msmarco-specb-bitfit

Usage

For usage instructions, refer to our codebase: https://github.com/Muennighoff/sgpt

Evaluation Results

For eval results, refer to the eval folder or our paper: https://arxiv.org/abs/2202.08904

Training

The model was trained with the parameters:

DataLoader:

torch.utils.data.dataloader.DataLoader of length 124796 with parameters:

{'batch_size': 4, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}

Loss:

sentence_transformers.losses.MultipleNegativesRankingLoss.MultipleNegativesRankingLoss with parameters:

{'scale': 20.0, 'similarity_fct': 'cos_sim'}

Parameters of the fit()-Method:

{
    "epochs": 10,
    "evaluation_steps": 0,
    "evaluator": "NoneType",
    "max_grad_norm": 1,
    "optimizer_class": "<class 'transformers.optimization.AdamW'>",
    "optimizer_params": {
        "lr": 7.5e-05
    },
    "scheduler": "WarmupLinear",
    "steps_per_epoch": null,
    "warmup_steps": 1000,
    "weight_decay": 0.01
}

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 300, 'do_lower_case': False}) with Transformer model: GPTNeoModel 
  (1): Pooling({'word_embedding_dimension': 2560, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': True, 'pooling_mode_lasttoken': False})
)

Citing & Authors

@article{muennighoff2022sgpt,
  title={SGPT: GPT Sentence Embeddings for Semantic Search},
  author={Muennighoff, Niklas},
  journal={arXiv preprint arXiv:2202.08904},
  year={2022}
}

Downloads last month: 269

Hosted inference API

Sentence Similarity

Examples

This model can be loaded on the Inference API on-demand.

Evaluation results

accuracy on MTEB AmazonCounterfactualClassification (en)
test set self-reported

67.567
ap on MTEB AmazonCounterfactualClassification (en)
test set self-reported

30.756
f1 on MTEB AmazonCounterfactualClassification (en)
test set self-reported

61.805
accuracy on MTEB AmazonPolarityClassification
test set self-reported

71.440
ap on MTEB AmazonPolarityClassification
test set self-reported

65.913
f1 on MTEB AmazonPolarityClassification
test set self-reported

70.906
accuracy on MTEB AmazonReviewsClassification (en)
test set self-reported

35.748
f1 on MTEB AmazonReviewsClassification (en)
test set self-reported

35.486
map_at_1 on MTEB ArguAna
test set self-reported

25.960
map_at_10 on MTEB ArguAna
test set self-reported

41.619
map_at_100 on MTEB ArguAna
test set self-reported

42.673
map_at_1000 on MTEB ArguAna
test set self-reported

42.684
map_at_3 on MTEB ArguAna
test set self-reported

36.569
map_at_5 on MTEB ArguAna
test set self-reported

39.397
mrr_at_1 on MTEB ArguAna
test set self-reported

26.316
mrr_at_10 on MTEB ArguAna
test set self-reported

41.772
mrr_at_100 on MTEB ArguAna
test set self-reported

42.820
mrr_at_1000 on MTEB ArguAna
test set self-reported

42.830
mrr_at_3 on MTEB ArguAna
test set self-reported

36.724
mrr_at_5 on MTEB ArguAna
test set self-reported

39.529
ndcg_at_1 on MTEB ArguAna
test set self-reported

25.960
ndcg_at_10 on MTEB ArguAna
test set self-reported

50.491
ndcg_at_100 on MTEB ArguAna
test set self-reported

54.865
ndcg_at_1000 on MTEB ArguAna
test set self-reported

55.107
ndcg_at_3 on MTEB ArguAna
test set self-reported

40.053
ndcg_at_5 on MTEB ArguAna
test set self-reported

45.134
precision_at_1 on MTEB ArguAna
test set self-reported

25.960
precision_at_10 on MTEB ArguAna
test set self-reported

7.895
precision_at_100 on MTEB ArguAna
test set self-reported

0.978
precision_at_1000 on MTEB ArguAna
test set self-reported

0.100
precision_at_3 on MTEB ArguAna
test set self-reported

16.714
precision_at_5 on MTEB ArguAna
test set self-reported

12.489
recall_at_1 on MTEB ArguAna
test set self-reported

25.960
recall_at_10 on MTEB ArguAna
test set self-reported

78.947
recall_at_100 on MTEB ArguAna
test set self-reported

97.795
recall_at_1000 on MTEB ArguAna
test set self-reported

99.644
recall_at_3 on MTEB ArguAna
test set self-reported

50.142
recall_at_5 on MTEB ArguAna
test set self-reported

62.447
v_measure on MTEB ArxivClusteringP2P
test set self-reported

44.721
v_measure on MTEB ArxivClusteringS2S
test set self-reported

35.081
map on MTEB AskUbuntuDupQuestions
test set self-reported

59.635
mrr on MTEB AskUbuntuDupQuestions
test set self-reported

73.681
cos_sim_pearson on MTEB BIOSSES
test set self-reported

87.428
cos_sim_spearman on MTEB BIOSSES
test set self-reported

84.843
euclidean_pearson on MTEB BIOSSES
test set self-reported

85.593
euclidean_spearman on MTEB BIOSSES
test set self-reported

85.853
manhattan_pearson on MTEB BIOSSES
test set self-reported

85.412
manhattan_spearman on MTEB BIOSSES
test set self-reported

85.523
accuracy on MTEB Banking77Classification
test set self-reported

83.218
f1 on MTEB Banking77Classification
test set self-reported

83.154
v_measure on MTEB BiorxivClusteringP2P
test set self-reported

34.414

View leaderboard (Papers With Code)