README.md

metadata

base_model: liddlefish/privacy_embedding_rag_10k_base_checkpoint_2
language:
  - en
library_name: sentence-transformers
license: apache-2.0
metrics:
  - cosine_accuracy@1
  - cosine_accuracy@3
  - cosine_accuracy@5
  - cosine_accuracy@10
  - cosine_precision@1
  - cosine_precision@3
  - cosine_precision@5
  - cosine_precision@10
  - cosine_recall@1
  - cosine_recall@3
  - cosine_recall@5
  - cosine_recall@10
  - cosine_ndcg@10
  - cosine_mrr@10
  - cosine_map@100
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - dataset_size:1K<n<10K
  - loss:MatryoshkaLoss
  - loss:MultipleNegativesRankingLoss
widget:
  - source_sentence: Żywot św. Stanisława
    sentences:
      - czym różni się Żywot św. Stanisława od Legendy św. Stanisława?
      - czemu gra The Saboteur wywołała wiele kontrowersji?
      - Muzykę do obrazu skomponowali Marco Frisina i Ennio Morricone.
  - source_sentence: Jaakow Jicchak Szapira
    sentences:
      - cadykiem którego miasta był Jaakow Jicchak Dan Landau?
      - gdzie zginął przedwojenny minister Antoni Olszewski?
      - ' W 1867 oddano do użytku Kolej Warszawsko-Terespolską (całą linię).'
  - source_sentence: Chłopiec z Nariokotome
    sentences:
      - ile wynosiła objętość mózgu chłopca z Nariokotome?
      - czemu gra The Saboteur wywołała wiele kontrowersji?
      - Akcja powieści rozgrywa się w XV-wiecznej Polsce.
  - source_sentence: Stany Zjednoczone Polski
    sentences:
      - kiedy miały szansę powstać Stany Zjednoczone Polski?
      - z jakiego powodu Chloé wywołała skandal w Melbourne?
      - komu przysługiwał tytuł autokratora?
  - source_sentence: Sen o zastrzyku Irmy
    sentences:
      - gdzie Freud spotkał Irmę we śnie o zastrzyku Irmy?
      - dlaczego Ōkunoshima została wymazana z map Japonii?
      - który samochód wyglądem nawiązuje do Mercedesa-Benza SLS AMG?
model-index:
  - name: privacy_embedding_rag_10k_base_checkpoint_2-klej-dyk-v0.1
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 768
          type: dim_768
        metrics:
          - type: cosine_accuracy@1
            value: 0.1875
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.4543269230769231
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.6225961538461539
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.7307692307692307
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.1875
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.15144230769230768
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.12451923076923076
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.07307692307692307
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.1875
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.4543269230769231
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.6225961538461539
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.7307692307692307
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.4453345212200682
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.35500896672771654
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.36239083059244687
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 512
          type: dim_512
        metrics:
          - type: cosine_accuracy@1
            value: 0.18269230769230768
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.44471153846153844
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.6033653846153846
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.7115384615384616
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.18269230769230768
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.14823717948717946
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.12067307692307691
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.07115384615384614
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.18269230769230768
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.44471153846153844
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.6033653846153846
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.7115384615384616
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.43488982498130374
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.347151633089133
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.3548109777991144
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 256
          type: dim_256
        metrics:
          - type: cosine_accuracy@1
            value: 0.1875
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.4230769230769231
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.5576923076923077
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.6682692307692307
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.1875
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.14102564102564102
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.11153846153846153
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.06682692307692308
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.1875
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.4230769230769231
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.5576923076923077
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.6682692307692307
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.41398239515933494
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.3336862789987789
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.3427233175204077
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 128
          type: dim_128
        metrics:
          - type: cosine_accuracy@1
            value: 0.17067307692307693
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.36778846153846156
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.5120192307692307
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.6009615384615384
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.17067307692307693
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.12259615384615384
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.10240384615384614
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.06009615384615385
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.17067307692307693
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.36778846153846156
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.5120192307692307
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.6009615384615384
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.371201964014572
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.2987818605006104
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.3066873839005868
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 64
          type: dim_64
        metrics:
          - type: cosine_accuracy@1
            value: 0.15865384615384615
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.31009615384615385
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.3870192307692308
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.49278846153846156
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.15865384615384615
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.10336538461538461
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.07740384615384616
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.04927884615384615
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.15865384615384615
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.31009615384615385
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.3870192307692308
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.49278846153846156
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.3130531482964966
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.2569225045787546
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.2651139739879426
            name: Cosine Map@100

privacy_embedding_rag_10k_base_checkpoint_2-klej-dyk-v0.1

This is a sentence-transformers model finetuned from liddlefish/privacy_embedding_rag_10k_base_checkpoint_2. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: liddlefish/privacy_embedding_rag_10k_base_checkpoint_2
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity
• Language: en
• License: apache-2.0

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'Sen o zastrzyku Irmy',
    'gdzie Freud spotkał Irmę we śnie o zastrzyku Irmy?',
    'dlaczego Ōkunoshima została wymazana z map Japonii?',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

• Dataset: dim_768
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1875
cosine_accuracy@3	0.4543
cosine_accuracy@5	0.6226
cosine_accuracy@10	0.7308
cosine_precision@1	0.1875
cosine_precision@3	0.1514
cosine_precision@5	0.1245
cosine_precision@10	0.0731
cosine_recall@1	0.1875
cosine_recall@3	0.4543
cosine_recall@5	0.6226
cosine_recall@10	0.7308
cosine_ndcg@10	0.4453
cosine_mrr@10	0.355
cosine_map@100	0.3624

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1827
cosine_accuracy@3	0.4447
cosine_accuracy@5	0.6034
cosine_accuracy@10	0.7115
cosine_precision@1	0.1827
cosine_precision@3	0.1482
cosine_precision@5	0.1207
cosine_precision@10	0.0712
cosine_recall@1	0.1827
cosine_recall@3	0.4447
cosine_recall@5	0.6034
cosine_recall@10	0.7115
cosine_ndcg@10	0.4349
cosine_mrr@10	0.3472
cosine_map@100	0.3548

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1875
cosine_accuracy@3	0.4231
cosine_accuracy@5	0.5577
cosine_accuracy@10	0.6683
cosine_precision@1	0.1875
cosine_precision@3	0.141
cosine_precision@5	0.1115
cosine_precision@10	0.0668
cosine_recall@1	0.1875
cosine_recall@3	0.4231
cosine_recall@5	0.5577
cosine_recall@10	0.6683
cosine_ndcg@10	0.414
cosine_mrr@10	0.3337
cosine_map@100	0.3427

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1707
cosine_accuracy@3	0.3678
cosine_accuracy@5	0.512
cosine_accuracy@10	0.601
cosine_precision@1	0.1707
cosine_precision@3	0.1226
cosine_precision@5	0.1024
cosine_precision@10	0.0601
cosine_recall@1	0.1707
cosine_recall@3	0.3678
cosine_recall@5	0.512
cosine_recall@10	0.601
cosine_ndcg@10	0.3712
cosine_mrr@10	0.2988
cosine_map@100	0.3067

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1587
cosine_accuracy@3	0.3101
cosine_accuracy@5	0.387
cosine_accuracy@10	0.4928
cosine_precision@1	0.1587
cosine_precision@3	0.1034
cosine_precision@5	0.0774
cosine_precision@10	0.0493
cosine_recall@1	0.1587
cosine_recall@3	0.3101
cosine_recall@5	0.387
cosine_recall@10	0.4928
cosine_ndcg@10	0.3131
cosine_mrr@10	0.2569
cosine_map@100	0.2651

Training Details

Training Dataset

Unnamed Dataset

• Size: 3,738 training samples
• Columns: positive and anchor
• Approximate statistics based on the first 1000 samples:

  	positive	anchor
  type	string	string
  details	min: 7 tokens mean: 89.43 tokens max: 507 tokens	min: 9 tokens mean: 30.98 tokens max: 76 tokens

• Samples:

  positive	anchor
  Zespół Blaua (zespół Jabsa, ang. Blau syndrome, BS) – rzadka choroba genetyczna o dziedziczeniu autosomalnym dominującym, charakteryzująca się ziarniniakowym zapaleniem stawów o wczesnym początku, zapaleniem jagodówki (uveitis) i wysypką skórną, a także kamptodaktylią.	jakie choroby genetyczne dziedziczą się autosomalnie dominująco?
  Gorgippia Gorgippia – starożytne miasto bosporańskie nad Morzem Czarnym, którego pozostałości znajdują się obecnie pod współczesną zabudową centralnej części miasta Anapa w Kraju Krasnodarskim w Rosji.	gdzie obecnie znajduje się starożytne miasto Gorgippia?
  Ulubionym dystansem Rücker było 400 metrów i to na nim notowała największe indywidualne sukcesy : srebrny medal Mistrzostw Europy juniorów w lekkoatletyce (Saloniki 1991) 6. miejsce w Pucharze Świata w Lekkoatletyce (Hawana 1992) 5. miejsce na Mistrzostwach Europy w Lekkoatletyce (Helsinki 1994) srebro podczas Mistrzostw Świata w Lekkoatletyce (Sewilla 1999) złota medalistka mistrzostw Niemiec Duże sukcesy odnosiła także w sztafecie 4 x 400 metrów : złoto Mistrzostw Europy juniorów w lekkoatletyce (Varaždin 1989) złoty medal Mistrzostw Europy juniorów w lekkoatletyce (Saloniki 1991) brąz na Mistrzostwach Europy w Lekkoatletyce (Helsinki 1994) brązowy medal podczas Igrzysk Olimpijskich (Atlanta 1996) brąz na Halowych Mistrzostwach Świata w Lekkoatletyce (Paryż 1997) złoto Mistrzostw Świata w Lekkoatletyce (Ateny 1997) brązowy medal Mistrzostw Świata w Lekkoatletyce (Sewilla 1999)	kto zaprojektował medale, które będą wręczane podczas tegorocznych mistrzostw Europy juniorów w lekkoatletyce?

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 5
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• bf16: True
• tf32: True
• load_best_model_at_end: True
• optim: adamw_torch_fused
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• eval_accumulation_steps: None
• learning_rate: 2e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 5
• max_steps: -1
• lr_scheduler_type: cosine
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: True
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 42
• data_seed: None
• jit_mode_eval: False
• use_ipex: False
• bf16: True
• fp16: False
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: True
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: True
• ignore_data_skip: False
• fsdp: []
• fsdp_min_num_params: 0
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• fsdp_transformer_layer_cls_to_wrap: None
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: False
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• eval_do_concat_batches: True
• fp16_backend: auto
• push_to_hub_model_id: None
• push_to_hub_organization: None
• mp_parameters:
• auto_find_batch_size: False
• full_determinism: False
• torchdynamo: None
• ray_scope: last
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• dispatch_batches: None
• split_batches: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	dim_128_cosine_map@100	dim_256_cosine_map@100	dim_512_cosine_map@100	dim_64_cosine_map@100	dim_768_cosine_map@100
0.0684	1	9.112	-	-	-	-	-
0.1368	2	9.5133	-	-	-	-	-
0.2051	3	9.0946	-	-	-	-	-
0.2735	4	8.9744	-	-	-	-	-
0.3419	5	7.9039	-	-	-	-	-
0.4103	6	8.1973	-	-	-	-	-
0.4786	7	6.8979	-	-	-	-	-
0.5470	8	7.0324	-	-	-	-	-
0.6154	9	6.6472	-	-	-	-	-
0.6838	10	6.3009	-	-	-	-	-
0.7521	11	6.8778	-	-	-	-	-
0.8205	12	5.9809	-	-	-	-	-
0.8889	13	5.3054	-	-	-	-	-
0.9573	14	5.706	0.2868	0.3280	0.3522	0.2415	0.3477
1.0256	15	5.0592	-	-	-	-	-
1.0940	16	4.7655	-	-	-	-	-
1.1624	17	4.9682	-	-	-	-	-
1.2308	18	5.1226	-	-	-	-	-
1.2991	19	4.8655	-	-	-	-	-
1.3675	20	4.2008	-	-	-	-	-
1.4359	21	5.0281	-	-	-	-	-
1.5043	22	4.3074	-	-	-	-	-
1.5726	23	4.3163	-	-	-	-	-
1.6410	24	3.9344	-	-	-	-	-
1.7094	25	4.6567	-	-	-	-	-
1.7778	26	4.5145	-	-	-	-	-
1.8462	27	4.1319	-	-	-	-	-
1.9145	28	3.8768	-	-	-	-	-
1.9829	29	3.5525	0.2986	0.3330	0.3483	0.2590	0.3534
2.0513	30	3.8693	-	-	-	-	-
2.1197	31	3.4675	-	-	-	-	-
2.1880	32	4.0598	-	-	-	-	-
2.2564	33	4.2429	-	-	-	-	-
2.3248	34	3.3686	-	-	-	-	-
2.3932	35	3.2663	-	-	-	-	-
2.4615	36	3.8585	-	-	-	-	-
2.5299	37	3.1157	-	-	-	-	-
2.5983	38	3.5254	-	-	-	-	-
2.6667	39	3.2782	-	-	-	-	-
2.7350	40	4.3151	-	-	-	-	-
2.8034	41	3.4567	-	-	-	-	-
2.8718	42	3.3976	-	-	-	-	-
2.9402	43	3.3945	0.3014	0.3343	0.3522	0.2626	0.3593
3.0085	44	3.4487	-	-	-	-	-
3.0769	45	3.0021	-	-	-	-	-
3.1453	46	3.2332	-	-	-	-	-
3.2137	47	3.3012	-	-	-	-	-
3.2821	48	3.2735	-	-	-	-	-
3.3504	49	2.5335	-	-	-	-	-
3.4188	50	3.7025	-	-	-	-	-
3.4872	51	2.8596	-	-	-	-	-
3.5556	52	3.1108	-	-	-	-	-
3.6239	53	3.2807	-	-	-	-	-
3.6923	54	3.1604	-	-	-	-	-
3.7607	55	3.7179	-	-	-	-	-
3.8291	56	3.3418	-	-	-	-	-
3.8974	57	2.9735	-	-	-	-	-
3.9658	58	3.2755	0.3066	0.3409	0.3546	0.2653	0.3626
4.0342	59	3.1444	-	-	-	-	-
4.1026	60	3.0212	-	-	-	-	-
4.1709	61	3.1298	-	-	-	-	-
4.2393	62	3.3195	-	-	-	-	-
4.3077	63	2.996	-	-	-	-	-
4.3761	64	2.4636	-	-	-	-	-
4.4444	65	3.2388	-	-	-	-	-
4.5128	66	2.747	-	-	-	-	-
4.5812	67	2.8715	-	-	-	-	-
4.6496	68	3.1402	-	-	-	-	-
4.7179	69	3.547	-	-	-	-	-
4.7863	70	3.6094	0.3067	0.3427	0.3548	0.2651	0.3624

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.12.2
• Sentence Transformers: 3.0.0
• Transformers: 4.41.2
• PyTorch: 2.3.1
• Accelerate: 0.27.2
• Datasets: 2.19.1
• Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning}, 
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply}, 
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}