README.md

metadata

base_model: BAAI/bge-m3
library_name: sentence-transformers
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
  - generated_from_trainer
  - dataset_size:8769
  - loss:MatryoshkaLoss
  - loss:MultipleNegativesRankingLoss
widget:
  - source_sentence: >-
      Aquelles persones que fan un ús regular i continuat de la deixalleria
      municipal poden gaudir d’una bonificació del 20% sobre la quota de les
      taxes per recollida, tractament i eliminació d'escombraries i altres
      residus urbans domiciliaris.
    sentences:
      - Quin és el contingut dels documents dirigits a l'Ajuntament de Sitges?
      - Quin és el benefici de la deixalleria municipal?
      - Quin és el mètode de pagament dels ajuts atorgats en cas de normalitat?
  - source_sentence: >-
      Les subvencions per al desenvolupament i/o consolidació de sectors
      econòmics del municipi tenen com a objectiu generar un benefici ambiental
      per al municipi, a través de la promoció de pràctiques sostenibles.
    sentences:
      - >-
        Quin és el requisit per a la llicència per a la modificació d'un règim
        de propietat horitzontal?
      - >-
        Quin és el benefici ambiental esperat de les subvencions per al
        desenvolupament i/o consolidació de sectors econòmics del municipi?
      - >-
        Quin és el propòsit de la liquidació de l'import corresponent a
        l'exercici?
  - source_sentence: >-
      Aquelles persones que s'hagin inscrit a les estades esportives
      organitzades per l'Ajuntament de Sitges i que formin part d'una unitat
      familiar amb uns ingressos bruts mensuals, que una vegada dividits pel
      nombre de membres, siguin inferiors entre una i dues terceres parts de
      l'IPREM, poden sol·licitar una reducció de la quota d'aquestes activitats
      o l'aplicació de la corresponent tarifa bonificada establerta en les
      ordenances dels preus públics.
    sentences:
      - Quin és el benefici de les subvencions per a les entitats culturals?
      - >-
        Quin és el paper de l'IPREM en la sol·licitud de reducció de la quota
        d'una estada esportiva?
      - >-
        Quin és el paper de l'Ajuntament en la resolució d'una situació
        sanitària no adequada en un domini particular?
  - source_sentence: >-
      La inscripció al cens municipal facilita la recuperació d’aquests animals
      en cas de pèrdua alhora que permet a l’Ajuntament disposar de les dades
      necessàries en cas que s’hagin de realitzar campanyes sanitàries.
    sentences:
      - >-
        Quin és el tipus de serveis auxiliars que es consideren despeses
        elegibles?
      - Quin és el benefici d'estacionar a les zones verdes per als residents?
      - >-
        Quin és el motiu pel qual es crea el cens municipal d’animals de
        companyia?
  - source_sentence: >-
      A la nostra vila hi ha veïns i veïnes que els agradaria tornar a fer de
      pagès o provar-ho per primera vegada.
    sentences:
      - >-
        Quin és l'objectiu principal de l'activitat del Viver dels Avis de
        Sitges?
      - Quin és el propòsit del carnet de conductor de taxi?
      - Quin és el paper de les persones en relació amb les indemnitzacions?
model-index:
  - name: SentenceTransformer based on BAAI/bge-m3
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 1024
          type: dim_1024
        metrics:
          - type: cosine_accuracy@1
            value: 0.11054852320675106
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.2270042194092827
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.30548523206751055
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.4531645569620253
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.11054852320675106
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.07566807313642755
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.06109704641350212
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.04531645569620253
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.11054852320675106
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.2270042194092827
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.30548523206751055
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.4531645569620253
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.25622764604771076
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.1965350612818966
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.21859411055862238
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 768
          type: dim_768
        metrics:
          - type: cosine_accuracy@1
            value: 0.11561181434599156
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.2320675105485232
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.31139240506329113
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.44556962025316454
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.11561181434599156
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.07735583684950773
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.06227848101265824
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.044556962025316456
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.11561181434599156
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.2320675105485232
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.31139240506329113
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.44556962025316454
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.2579660315889156
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.20086732301922164
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.22344331787470567
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 512
          type: dim_512
        metrics:
          - type: cosine_accuracy@1
            value: 0.10379746835443038
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.2210970464135021
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.2970464135021097
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.43966244725738396
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.10379746835443038
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.07369901547116735
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.05940928270042194
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.043966244725738395
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.10379746835443038
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.2210970464135021
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.2970464135021097
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.43966244725738396
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.2473619714740055
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.18892840399169497
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.21182552044674802
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 256
          type: dim_256
        metrics:
          - type: cosine_accuracy@1
            value: 0.10042194092827005
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.21518987341772153
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.2978902953586498
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.4438818565400844
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.10042194092827005
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.07172995780590716
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.05957805907172995
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.04438818565400844
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.10042194092827005
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.21518987341772153
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.2978902953586498
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.4438818565400844
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.2479637375723138
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.18831156653941447
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.21130848497160895
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 128
          type: dim_128
        metrics:
          - type: cosine_accuracy@1
            value: 0.10886075949367088
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.22616033755274262
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.3029535864978903
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.4413502109704641
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.10886075949367088
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.07538677918424753
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.060590717299578066
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.04413502109704641
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.10886075949367088
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.22616033755274262
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.3029535864978903
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.4413502109704641
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.25366131313332974
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.19639441430580665
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.2187767008895725
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 64
          type: dim_64
        metrics:
          - type: cosine_accuracy@1
            value: 0.09367088607594937
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.2
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.2742616033755274
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.4177215189873418
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.09367088607594937
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.06666666666666667
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.05485232067510549
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.04177215189873418
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.09367088607594937
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.2
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.2742616033755274
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.4177215189873418
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.23046340016141767
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.1738279418659165
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.19782551958501599
            name: Cosine Map@100

SentenceTransformer based on BAAI/bge-m3

This is a sentence-transformers model finetuned from BAAI/bge-m3 on the json dataset. It maps sentences & paragraphs to a 1024-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: BAAI/bge-m3
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 1024 tokens
• Similarity Function: Cosine Similarity
• Training Dataset:
  • json

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': 8192, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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("adriansanz/ST-tramits-sitges-003-5ep")
# Run inference
sentences = [
    'A la nostra vila hi ha veïns i veïnes que els agradaria tornar a fer de pagès o provar-ho per primera vegada.',
    "Quin és l'objectiu principal de l'activitat del Viver dels Avis de Sitges?",
    'Quin és el paper de les persones en relació amb les indemnitzacions?',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Evaluation

Metrics

Information Retrieval

• Dataset: dim_1024
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1105
cosine_accuracy@3	0.227
cosine_accuracy@5	0.3055
cosine_accuracy@10	0.4532
cosine_precision@1	0.1105
cosine_precision@3	0.0757
cosine_precision@5	0.0611
cosine_precision@10	0.0453
cosine_recall@1	0.1105
cosine_recall@3	0.227
cosine_recall@5	0.3055
cosine_recall@10	0.4532
cosine_ndcg@10	0.2562
cosine_mrr@10	0.1965
cosine_map@100	0.2186

Information Retrieval

• Dataset: dim_768
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1156
cosine_accuracy@3	0.2321
cosine_accuracy@5	0.3114
cosine_accuracy@10	0.4456
cosine_precision@1	0.1156
cosine_precision@3	0.0774
cosine_precision@5	0.0623
cosine_precision@10	0.0446
cosine_recall@1	0.1156
cosine_recall@3	0.2321
cosine_recall@5	0.3114
cosine_recall@10	0.4456
cosine_ndcg@10	0.258
cosine_mrr@10	0.2009
cosine_map@100	0.2234

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1038
cosine_accuracy@3	0.2211
cosine_accuracy@5	0.297
cosine_accuracy@10	0.4397
cosine_precision@1	0.1038
cosine_precision@3	0.0737
cosine_precision@5	0.0594
cosine_precision@10	0.044
cosine_recall@1	0.1038
cosine_recall@3	0.2211
cosine_recall@5	0.297
cosine_recall@10	0.4397
cosine_ndcg@10	0.2474
cosine_mrr@10	0.1889
cosine_map@100	0.2118

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1004
cosine_accuracy@3	0.2152
cosine_accuracy@5	0.2979
cosine_accuracy@10	0.4439
cosine_precision@1	0.1004
cosine_precision@3	0.0717
cosine_precision@5	0.0596
cosine_precision@10	0.0444
cosine_recall@1	0.1004
cosine_recall@3	0.2152
cosine_recall@5	0.2979
cosine_recall@10	0.4439
cosine_ndcg@10	0.248
cosine_mrr@10	0.1883
cosine_map@100	0.2113

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1089
cosine_accuracy@3	0.2262
cosine_accuracy@5	0.303
cosine_accuracy@10	0.4414
cosine_precision@1	0.1089
cosine_precision@3	0.0754
cosine_precision@5	0.0606
cosine_precision@10	0.0441
cosine_recall@1	0.1089
cosine_recall@3	0.2262
cosine_recall@5	0.303
cosine_recall@10	0.4414
cosine_ndcg@10	0.2537
cosine_mrr@10	0.1964
cosine_map@100	0.2188

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0937
cosine_accuracy@3	0.2
cosine_accuracy@5	0.2743
cosine_accuracy@10	0.4177
cosine_precision@1	0.0937
cosine_precision@3	0.0667
cosine_precision@5	0.0549
cosine_precision@10	0.0418
cosine_recall@1	0.0937
cosine_recall@3	0.2
cosine_recall@5	0.2743
cosine_recall@10	0.4177
cosine_ndcg@10	0.2305
cosine_mrr@10	0.1738
cosine_map@100	0.1978

Training Details

Training Dataset

json

• Dataset: json
• Size: 8,769 training samples
• Columns: positive and anchor
• Approximate statistics based on the first 1000 samples:

  	positive	anchor
  type	string	string
  details	min: 5 tokens mean: 49.22 tokens max: 178 tokens	min: 10 tokens mean: 20.94 tokens max: 48 tokens

• Samples:

  positive	anchor
  L'Ajuntament de Sitges atorga subvencions per a projectes i activitats d'interès públic o social que tinguin per finalitat les activitats esportives federades, escolars o populars desenvolupades per les entitats esportives i esportistes del municipi de Sitges.	Quin és el benefici de les subvencions per a les entitats esportives?
  L'Ajuntament de Sitges atorga subvencions per a projectes i activitats d'interès públic o social que tinguin per finalitat les activitats esportives federades, escolars o populars desenvolupades per les entitats esportives i esportistes del municipi de Sitges al llarg de l'exercici per la qual es sol·licita la subvenció, i reuneixin les condicions assenyalades a les bases.	Quin és el període d'execució dels projectes i activitats esportives?
  Certificat on s'indica el nombre d'habitatges que configuren el padró de l'Impost sobre Béns Immobles del municipi o bé d'una part d'aquest.	Quin és el contingut del certificat del nombre d'habitatges?

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          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: 0.2
• 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
• torch_empty_cache_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: 0.2
• 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
• eval_on_start: False
• eval_use_gather_object: False
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	dim_1024_cosine_map@100	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
0.2914	10	3.6318	-	-	-	-	-	-
0.5829	20	2.329	-	-	-	-	-	-
0.8743	30	1.5614	-	-	-	-	-	-
0.9909	34	-	0.2055	0.1998	0.2020	0.2001	0.1903	0.2019
1.1658	40	1.2383	-	-	-	-	-	-
1.4572	50	0.9323	-	-	-	-	-	-
1.7486	60	0.6616	-	-	-	-	-	-
1.9818	68	-	0.2244	0.2063	0.2223	0.2166	0.2011	0.2235
2.0401	70	0.5545	-	-	-	-	-	-
2.3315	80	0.5043	-	-	-	-	-	-
2.6230	90	0.3542	-	-	-	-	-	-
2.9144	100	0.3095	-	-	-	-	-	-
2.9727	102	-	0.2224	0.2046	0.2170	0.2100	0.1986	0.2144
3.2058	110	0.2863	-	-	-	-	-	-
3.4973	120	0.2329	-	-	-	-	-	-
3.7887	130	0.2353	-	-	-	-	-	-
3.9927	137	-	0.2197	0.2112	0.2098	0.2154	0.1949	0.2178
4.0801	140	0.1759	-	-	-	-	-	-
4.3716	150	0.2308	-	-	-	-	-	-
4.6630	160	0.1656	-	-	-	-	-	-
4.9545	170	0.1812	0.2186	0.2188	0.2113	0.2118	0.1978	0.2234

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.1.1
• Transformers: 4.44.2
• PyTorch: 2.4.1+cu121
• Accelerate: 0.35.0.dev0
• Datasets: 3.0.1
• 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}
}