fabskill/job_and_title_siamese_binary

SentenceTransformer based on sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2

This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2. It maps sentences & paragraphs to a 384-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: sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 384 tokens
• Similarity Function: Cosine Similarity

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': 128, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

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("training_job_matching_sentence-transformers-paraphrase-multilingual-MiniLM-L12-v2-2024-09-03_13-14-25")
# Run inference
sentences = [
    "Responsable d'élevage en production ovine",
    "de gestion immobilière estimer la valeur d'un bien, d'un produit droit, contentieux et négociationappliquer un cadre juridique ou réglementaire réaliser le suivi des décisions prises en assemblées de copropriété traiter des dossiers de contentieux réaliser la gestion administrative des contrats management animer, coordonner une équipe gestion des ressources humaines gérer les ressources humaines conseil, transmission assurer une médiation constructionétablir l'état d'avancement de travaux piloter la préparation de travaux planifier des travaux de rénovation définir les besoins en rénovation du patrimoine immobilier",
    "de travail et risques professionnelsau domicile d'un particulier déplacements professionnels port d'équipement de protection individuel (epi) : gants, chaussures, casque, protections auditives horaires et durée du travailtravail en astreinte travail le week-end publics spécifiques particuliers secteurs d'activité • bâtiment et travaux publics (btp) 4 / 4 -",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

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

Evaluation

Metrics

Binary Classification

• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.9819
cosine_accuracy_threshold	0.6798
cosine_f1	0.9734
cosine_f1_threshold	0.6782
cosine_precision	0.9712
cosine_recall	0.9756
cosine_ap	0.9778
dot_accuracy	0.9799
dot_accuracy_threshold	169.4876
dot_f1	0.9706
dot_f1_threshold	169.4876
dot_precision	0.9627
dot_recall	0.9786
dot_ap	0.9774
manhattan_accuracy	0.9756
manhattan_accuracy_threshold	160.5027
manhattan_f1	0.9638
manhattan_f1_threshold	165.2382
manhattan_precision	0.9673
manhattan_recall	0.9603
manhattan_ap	0.9782
euclidean_accuracy	0.9827
euclidean_accuracy_threshold	12.7983
euclidean_f1	0.9745
euclidean_f1_threshold	12.8575
euclidean_precision	0.9733
euclidean_recall	0.9756
euclidean_ap	0.9783
max_accuracy	0.9827
max_accuracy_threshold	169.4876
max_f1	0.9745
max_f1_threshold	169.4876
max_precision	0.9733
max_recall	0.9786
max_ap	0.9783

Binary Classification

• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.8349
cosine_accuracy_threshold	0.9927
cosine_f1	0.5193
cosine_f1_threshold	0.7801
cosine_precision	0.4292
cosine_recall	0.6573
cosine_ap	0.5436
dot_accuracy	0.8377
dot_accuracy_threshold	247.4402
dot_f1	0.5101
dot_f1_threshold	180.7264
dot_precision	0.3992
dot_recall	0.7063
dot_ap	0.5302
manhattan_accuracy	0.8363
manhattan_accuracy_threshold	24.4719
manhattan_f1	0.5027
manhattan_f1_threshold	122.6577
manhattan_precision	0.4097
manhattan_recall	0.6503
manhattan_ap	0.5317
euclidean_accuracy	0.8363
euclidean_accuracy_threshold	1.9895
euclidean_f1	0.5251
euclidean_f1_threshold	10.4537
euclidean_precision	0.4372
euclidean_recall	0.6573
euclidean_ap	0.544
max_accuracy	0.8377
max_accuracy_threshold	247.4402
max_f1	0.5251
max_f1_threshold	180.7264
max_precision	0.4372
max_recall	0.7063
max_ap	0.544

Binary Classification

• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.91
cosine_accuracy_threshold	0.8936
cosine_f1	0.7556
cosine_f1_threshold	0.7639
cosine_precision	0.8031
cosine_recall	0.7133
cosine_ap	0.7999
dot_accuracy	0.9127
dot_accuracy_threshold	227.503
dot_f1	0.7576
dot_f1_threshold	227.503
dot_precision	0.8264
dot_recall	0.6993
dot_ap	0.7881
manhattan_accuracy	0.9113
manhattan_accuracy_threshold	109.2699
manhattan_f1	0.7556
manhattan_f1_threshold	121.613
manhattan_precision	0.8031
manhattan_recall	0.7133
manhattan_ap	0.7969
euclidean_accuracy	0.91
euclidean_accuracy_threshold	7.6809
euclidean_f1	0.7556
euclidean_f1_threshold	11.5803
euclidean_precision	0.8031
euclidean_recall	0.7133
euclidean_ap	0.8007
max_accuracy	0.9127
max_accuracy_threshold	227.503
max_f1	0.7576
max_f1_threshold	227.503
max_precision	0.8264
max_recall	0.7133
max_ap	0.8007

Binary Classification

• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.8809
cosine_accuracy_threshold	0.7636
cosine_f1	0.7021
cosine_f1_threshold	0.553
cosine_precision	0.6819
cosine_recall	0.7236
cosine_ap	0.7361
dot_accuracy	0.8787
dot_accuracy_threshold	217.5387
dot_f1	0.7004
dot_f1_threshold	164.1041
dot_precision	0.7004
dot_recall	0.7004
dot_ap	0.7299
manhattan_accuracy	0.8782
manhattan_accuracy_threshold	146.0133
manhattan_f1	0.7016
manhattan_f1_threshold	180.2034
manhattan_precision	0.6847
manhattan_recall	0.7194
manhattan_ap	0.7262
euclidean_accuracy	0.8804
euclidean_accuracy_threshold	13.7647
euclidean_f1	0.7046
euclidean_f1_threshold	15.2429
euclidean_precision	0.7046
euclidean_recall	0.7046
euclidean_ap	0.7391
max_accuracy	0.8809
max_accuracy_threshold	217.5387
max_f1	0.7046
max_f1_threshold	180.2034
max_precision	0.7046
max_recall	0.7236
max_ap	0.7391

Binary Classification

• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.9316
cosine_accuracy_threshold	0.63
cosine_f1	0.8316
cosine_f1_threshold	0.5285
cosine_precision	0.7843
cosine_recall	0.885
cosine_ap	0.8867
dot_accuracy	0.9293
dot_accuracy_threshold	199.23
dot_f1	0.8274
dot_f1_threshold	165.8962
dot_precision	0.7892
dot_recall	0.8695
dot_ap	0.8867
manhattan_accuracy	0.9289
manhattan_accuracy_threshold	176.4425
manhattan_f1	0.821
manhattan_f1_threshold	176.4425
manhattan_precision	0.8303
manhattan_recall	0.8119
manhattan_ap	0.8726
euclidean_accuracy	0.932
euclidean_accuracy_threshold	14.7442
euclidean_f1	0.8337
euclidean_f1_threshold	16.6326
euclidean_precision	0.7952
euclidean_recall	0.8761
euclidean_ap	0.8886
max_accuracy	0.932
max_accuracy_threshold	199.23
max_f1	0.8337
max_f1_threshold	176.4425
max_precision	0.8303
max_recall	0.885
max_ap	0.8886

Training Details

Training Dataset

Unnamed Dataset

• Size: 42,735 training samples
• Columns: name, fiche, and label
• Approximate statistics based on the first 1000 samples:

  	name	fiche	label
  type	string	string	int
  details	min: 3 tokens mean: 9.44 tokens max: 44 tokens	min: 4 tokens mean: 107.63 tokens max: 128 tokens	0: ~78.20% 1: ~21.80%

• Samples:

  name	fiche	label
  Front End Angular Developer	communication WCF is used. The layer concept enables the reduction of dependencies (dependency injection) of the different tasks (separation of concerns). The entities are exchanged with the database via object-relational mapping (ORM) and processed using the CRUD methods. Through the consistent use of the MVVM pattern, we avoid code-behind. The user interface of the application is realized using the PRISM framework as a "Composite Application UI".Main tasks Software developerIn cooperation with a team located in Germany and respecting the software development guidelines and customers	0
  SCM : Administrateur des ventes	CHEF DE PROJET CONFIRME MAÎTRISANT ANGULAR 4.SON RÔLE SERA L'ENCADREMENT D'UNE EQUIPE ET LA GESTION TOTALE DU DÉVELOPPEMENT D'UNE APPLICATION MOBILE ANDROID.ESPRIT D'ÉQUIPE OBLIGATOIRE.	0
  Talent Acquisition Junior	Pilotage et suivi de toutes les activités du call center (commandes clients et interactions bénéficiaires de la carte).Assurer le calcul et le suivi des Kpi’s du call center.Veiller à la conformité des process et des procédures pour le call center.Contrôle de la prise en charge et la saisie des demandes et réclamations.Pilotage et suivi des projets de la direction clientèle.Assurer toutes demandes ou actions émanant de la Direction Clientèle.Assurer le maintien d’une bonne qualité de service.Augmenter la satisfaction	0

• Loss: SoftmaxLoss

Evaluation Dataset

Unnamed Dataset

• Size: 2,250 evaluation samples
• Columns: name, fiche, and label
• Approximate statistics based on the first 1000 samples:

  	name	fiche	label
  type	string	string	int
  details	min: 3 tokens mean: 9.31 tokens max: 36 tokens	min: 3 tokens mean: 109.04 tokens max: 128 tokens	0: ~80.30% 1: ~19.70%

• Samples:

  name	fiche	label
  1way com	Nous somme a la recherche de Profils en Telco avec connaissance en Produit! 😃Vous avez une connaissances dans la télécommunication? Emission ou réception (orange, sfr, boygues, free..)Vous voulez travailler dans un environnement stable, accueillant et sans pression?Vous êtes passionnés? Postulez maintenant et profitez d'un salaire motivant et pleins d'avantages:- Salaire 1100 a 1300 (selon le profil)- Primes et challenges- Tickets repas- Transport assuré- Samedi dimanche off- Titularisation- Convention	1
  Senior Front end Web Developer	As part of our growth in Tunis, we are looking to hire a Sénior Front-End Web Developer, who is passionate by Web Development and would like to have a career in an international company, in the Private Banking sector, within an exciting work environment.You will take part, throughout the software development life cycle (SDLC), to the requirement analysis, development and the support of different applications for private banks.You will perform AngularJS frontend development.You will integrate a highly motivated development team working on providing solutions for the private banking sector in which you will integrate the existing global	1
  DÉVELOPPEUR FULLSTACK RUBY ET ANGULAR	professionnel et d'évolution de carrière.- Projets stimulants et variés.- Esprit d'équipe et culture d'entreprise positive.- Salaire compétitif et avantages sociaux attractifs.Rejoignez MCOM et contribuez à révolutionner le commerce mobile avec nous!	0

• Loss: SoftmaxLoss

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 32
• num_train_epochs: 5
• warmup_ratio: 0.1
• bf16: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 32
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• learning_rate: 5e-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: linear
• 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: None
• 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: False
• 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
• 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
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss	loss	max_ap
0	0	-	-	0.6610
0.0673	300	0.638	-	-
0.1346	600	0.5642	-	-
0.2020	900	0.4942	-	-
0.2244	1000	-	0.4283	0.7756
0.2693	1200	0.4323	-	-
0.3366	1500	0.3986	-	-
0.4039	1800	0.3798	-	-
0.4488	2000	-	0.3481	0.8517
0.4713	2100	0.3532	-	-
0.5386	2400	0.3407	-	-
0.6059	2700	0.323	-	-
0.6732	3000	0.3022	0.2953	0.8899
0.7406	3300	0.2945	-	-
0.8079	3600	0.2864	-	-
0.8752	3900	0.2656	-	-
0.8977	4000	-	0.2434	0.9199
0.9425	4200	0.2581	-	-
1.0099	4500	0.2486	-	-
1.0772	4800	0.2282	-	-
1.1221	5000	-	0.2160	0.9248
1.1445	5100	0.2191	-	-
1.2118	5400	0.2113	-	-
1.2792	5700	0.2111	-	-
1.3465	6000	0.2011	0.1882	0.9339
1.4138	6300	0.1894	-	-
1.4811	6600	0.1814	-	-
1.5485	6900	0.1772	-	-
1.5709	7000	-	0.1697	0.9409
1.6158	7200	0.1731	-	-
1.6831	7500	0.1707	-	-
1.7504	7800	0.163	-	-
1.7953	8000	-	0.1497	0.9411
1.8178	8100	0.1576	-	-
1.8851	8400	0.1518	-	-
1.9524	8700	0.1447	-	-
2.0197	9000	0.142	0.1355	0.9483
2.0871	9300	0.1277	-	-
2.1544	9600	0.1278	-	-
2.2217	9900	0.1243	-	-
2.2442	10000	-	0.1225	0.9526
2.2890	10200	0.1228	-	-
2.3564	10500	0.1214	-	-
2.4237	10800	0.1173	-	-
2.4686	11000	-	0.1082	0.9606
2.4910	11100	0.1154	-	-
2.5583	11400	0.1098	-	-
2.6257	11700	0.1074	-	-
2.6930	12000	0.105	0.1005	0.9656
2.7603	12300	0.1042	-	-
2.8276	12600	0.0998	-	-
2.8950	12900	0.0967	-	-
2.9174	13000	-	0.0911	0.9645
2.9623	13200	0.0977	-	-
3.0296	13500	0.0896	-	-
3.0969	13800	0.0854	-	-
3.1418	14000	-	0.0843	0.9686
3.1643	14100	0.0848	-	-
3.2316	14400	0.0841	-	-
3.2989	14700	0.082	-	-
3.3662	15000	0.0815	0.0790	0.9711
3.4336	15300	0.0812	-	-
3.5009	15600	0.0799	-	-
3.5682	15900	0.0753	-	-
3.5907	16000	-	0.0751	0.9725
3.6355	16200	0.0756	-	-
3.7029	16500	0.0737	-	-
3.7702	16800	0.0742	-	-
3.8151	17000	-	0.0713	0.9750
3.8375	17100	0.0725	-	-
3.9048	17400	0.0721	-	-
3.9722	17700	0.0696	-	-
4.0395	18000	0.0665	0.0664	0.9746
4.1068	18300	0.0648	-	-
4.1741	18600	0.0636	-	-
4.2415	18900	0.0617	-	-
4.2639	19000	-	0.0637	0.9757
4.3088	19200	0.0624	-	-
4.3761	19500	0.062	-	-
4.4434	19800	0.0609	-	-
4.4883	20000	-	0.0608	0.9774
4.5108	20100	0.0607	-	-
4.5781	20400	0.061	-	-
4.6454	20700	0.0612	-	-
4.7127	21000	0.0598	0.0591	0.9777
4.7801	21300	0.0613	-	-
4.8474	21600	0.0599	-	-
4.9147	21900	0.0575	-	-
4.9372	22000	-	0.0582	0.9783
4.9820	22200	0.0593	-	-
5.0	22280	-	-	0.5440
0.8303	181	-	-	0.7148
0.4587	100	-	0.2849	0.7360
0.9174	200	-	0.3019	0.7230
1.3761	300	0.2712	0.2813	0.7697
1.8349	400	-	0.2667	0.8033
2.2936	500	-	0.2673	0.7936
2.7523	600	0.2268	0.2518	0.8078
3.2110	700	-	0.2539	0.8103
3.6697	800	-	0.2662	0.8118
4.1284	900	0.1845	0.2688	0.8003
4.5872	1000	-	0.2632	0.8081
0.4587	100	-	0.2642	0.8101
0.9174	200	-	0.2741	0.7995
1.3761	300	0.1742	0.2818	0.7861
1.8349	400	-	0.2595	0.8146
2.2936	500	-	0.2716	0.8021
2.7523	600	0.1572	0.2622	0.8013
3.2110	700	-	0.2660	0.7985
3.6697	800	-	0.2716	0.7986
4.1284	900	0.1327	0.2724	0.7942
4.5872	1000	-	0.2670	0.8007
5.0	1090	-	-	0.5292
0.1497	100	-	0.4254	0.5464
0.2994	200	-	0.3918	0.5718
0.4491	300	0.3988	0.3853	0.5670
0.5988	400	-	0.3670	0.5780
0.7485	500	-	0.3630	0.5954
0.8982	600	0.3577	0.3551	0.6197
1.0479	700	-	0.3463	0.6320
1.1976	800	-	0.3362	0.6455
1.3473	900	0.3092	0.3547	0.6496
1.4970	1000	-	0.3403	0.6502
1.6467	1100	-	0.3418	0.6614
1.7964	1200	0.2901	0.3367	0.6781
1.9461	1300	-	0.3283	0.6939
2.0958	1400	-	0.3266	0.7053
2.2455	1500	0.2627	0.3275	0.7074
2.3952	1600	-	0.3174	0.6976
2.5449	1700	-	0.3275	0.7037
2.6946	1800	0.2319	0.3094	0.7086
2.8443	1900	-	0.3184	0.7118
2.9940	2000	-	0.3195	0.7076
3.1437	2100	0.2222	0.3225	0.7178
3.2934	2200	-	0.3214	0.7184
3.4431	2300	-	0.3170	0.7270
3.5928	2400	0.188	0.3236	0.7269
3.7425	2500	-	0.3174	0.7345
3.8922	2600	-	0.3196	0.7365
4.0419	2700	0.1877	0.3174	0.7394
4.1916	2800	-	0.3195	0.7355
4.3413	2900	-	0.3207	0.7373
4.4910	3000	0.1582	0.3274	0.7349
4.6407	3100	-	0.3252	0.7350
4.7904	3200	-	0.3210	0.7393
4.9401	3300	0.1612	0.3205	0.7386
5.0	3340	-	-	0.8142
0.1497	100	-	0.2197	0.8248
0.2994	200	-	0.2117	0.8303
0.4491	300	0.2456	0.2299	0.8156
0.5988	400	-	0.2219	0.8113
0.7485	500	-	0.2149	0.8231
0.8982	600	0.2397	0.2110	0.8354
1.0479	700	-	0.2069	0.8479
1.1976	800	-	0.2070	0.8465
1.3473	900	0.1956	0.2046	0.8445
1.4970	1000	-	0.2070	0.8412
1.6467	1100	-	0.2001	0.8453
1.7964	1200	0.185	0.1970	0.8473
1.9461	1300	-	0.1904	0.8491
2.0958	1400	-	0.1864	0.8691
2.2455	1500	0.1537	0.1916	0.8570
2.3952	1600	-	0.1886	0.8740
2.5449	1700	-	0.1827	0.8770
2.6946	1800	0.1363	0.1771	0.8798
2.8443	1900	-	0.1768	0.8862
2.9940	2000	-	0.1799	0.8912
3.1437	2100	0.1276	0.1785	0.8838
3.2934	2200	-	0.1772	0.8803
3.4431	2300	-	0.1819	0.8801
3.5928	2400	0.1048	0.1763	0.8820
3.7425	2500	-	0.1782	0.8880
3.8922	2600	-	0.1784	0.8833
4.0419	2700	0.1017	0.1777	0.8885
4.1916	2800	-	0.1805	0.8901
4.3413	2900	-	0.1756	0.8911
4.4910	3000	0.0853	0.1781	0.8895
4.6407	3100	-	0.1784	0.8869
4.7904	3200	-	0.1775	0.8879
4.9401	3300	0.0854	0.1766	0.8883
5.0	3340	-	-	0.8886

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.42.4
• PyTorch: 2.3.1+cu121
• Accelerate: 0.32.1
• Datasets: 2.21.0
• Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers and SoftmaxLoss

@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",
}