--- 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:5214 - loss:MatryoshkaLoss - loss:MultipleNegativesRankingLoss widget: - source_sentence: Pel que fa als avals, la Junta de Govern Local en sessió celebrada el 4 de juliol de 2006, va aprovar els models d'aval en funció del concepte a garantir. sentences: - Quin és el benefici de la unitat de queixes i suggeriments per a la qualitat dels serveis de l'Ajuntament de Sitges? - Quin és el paper de la Junta de Govern Local? - Quin és el propòsit més important del tràmit de canvi de titular de la llicència de gual? - source_sentence: Per a tenir dret a ésser inscrit en el Registre de Sol·licitants d'Habitatge amb Protecció Oficial s'han de complir els procediments i els requisits establerts per normativa. sentences: - Quin és el paper de la persona sol·licitant en la gestió de les fiances o dipòsits d'una llicència d'obra? - Quin és el benefici de complir els procediments i els requisits establerts per normativa? - Quin és el centre cultural que es troba a l'Escorxador de Sitges i ofereix activitats culturals? - source_sentence: Aquest tràmit permet comunicar a l'Ajuntament de Sitges la finalització de les obres de nova construcció, o bé aquelles que hagin estat objecte de modificació substancial o d’ampliació quan per a l’autorització de les obres s’hagi exigit un projecte tècnic i a l’empara d’una llicència urbanística d’obra major. sentences: - Què passa si la modificació no té efectes sobre les persones o el medi ambient? - Quin és el requisit principal per a la gestió diària d'una colònia felina? - Quin és el paper del tràmit de comunicació prèvia de primera utilització i ocupació d'edificis i instal·lacions en el procés d'obtenció de la llicència urbanística d’obra major? - source_sentence: Es tracta dels ajuts per a la realització de la Inspecció Tècnica de l’Edifici (ITE) conjuntament amb l’elaboració dels certificats energètics. sentences: - Quins són els tipus de garanties que es poden ingressar? - Quin és el procés d’elaboració dels certificats energètics? - Quin és el paper de la consulta prèvia de classificació d'activitat en la tramitació administrativa municipal? - source_sentence: Les queixes, observacions i suggeriments són una eina important per a millorar la qualitat dels serveis municipals. sentences: - Quin és el propòsit dels ajuts econòmics? - Què és el que es busca amb les queixes, observacions i suggeriments? - Qui són les persones beneficiàries de l'ajut per a la creació de noves empreses? 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: 0.14367088607594936 name: Cosine Accuracy@1 - type: cosine_accuracy@3 value: 0.2818565400843882 name: Cosine Accuracy@3 - type: cosine_accuracy@5 value: 0.3930379746835443 name: Cosine Accuracy@5 - type: cosine_accuracy@10 value: 0.5664556962025317 name: Cosine Accuracy@10 - type: cosine_precision@1 value: 0.14367088607594936 name: Cosine Precision@1 - type: cosine_precision@3 value: 0.09395218002812938 name: Cosine Precision@3 - type: cosine_precision@5 value: 0.07860759493670887 name: Cosine Precision@5 - type: cosine_precision@10 value: 0.05664556962025316 name: Cosine Precision@10 - type: cosine_recall@1 value: 0.14367088607594936 name: Cosine Recall@1 - type: cosine_recall@3 value: 0.2818565400843882 name: Cosine Recall@3 - type: cosine_recall@5 value: 0.3930379746835443 name: Cosine Recall@5 - type: cosine_recall@10 value: 0.5664556962025317 name: Cosine Recall@10 - type: cosine_ndcg@10 value: 0.32426778614918705 name: Cosine Ndcg@10 - type: cosine_mrr@10 value: 0.25066212912731944 name: Cosine Mrr@10 - type: cosine_map@100 value: 0.2694799737895368 name: Cosine Map@100 - task: type: information-retrieval name: Information Retrieval dataset: name: dim 768 type: dim_768 metrics: - type: cosine_accuracy@1 value: 0.1470464135021097 name: Cosine Accuracy@1 - type: cosine_accuracy@3 value: 0.2871308016877637 name: Cosine Accuracy@3 - type: cosine_accuracy@5 value: 0.390084388185654 name: Cosine Accuracy@5 - type: cosine_accuracy@10 value: 0.5630801687763713 name: Cosine Accuracy@10 - type: cosine_precision@1 value: 0.1470464135021097 name: Cosine Precision@1 - type: cosine_precision@3 value: 0.09571026722925456 name: Cosine Precision@3 - type: cosine_precision@5 value: 0.07801687763713079 name: Cosine Precision@5 - type: cosine_precision@10 value: 0.056308016877637125 name: Cosine Precision@10 - type: cosine_recall@1 value: 0.1470464135021097 name: Cosine Recall@1 - type: cosine_recall@3 value: 0.2871308016877637 name: Cosine Recall@3 - type: cosine_recall@5 value: 0.390084388185654 name: Cosine Recall@5 - type: cosine_recall@10 value: 0.5630801687763713 name: Cosine Recall@10 - type: cosine_ndcg@10 value: 0.32549268557195893 name: Cosine Ndcg@10 - type: cosine_mrr@10 value: 0.25325421940928294 name: Cosine Mrr@10 - type: cosine_map@100 value: 0.272264774489146 name: Cosine Map@100 - task: type: information-retrieval name: Information Retrieval dataset: name: dim 512 type: dim_512 metrics: - type: cosine_accuracy@1 value: 0.14177215189873418 name: Cosine Accuracy@1 - type: cosine_accuracy@3 value: 0.28375527426160335 name: Cosine Accuracy@3 - type: cosine_accuracy@5 value: 0.3890295358649789 name: Cosine Accuracy@5 - type: cosine_accuracy@10 value: 0.5620253164556962 name: Cosine Accuracy@10 - type: cosine_precision@1 value: 0.14177215189873418 name: Cosine Precision@1 - type: cosine_precision@3 value: 0.09458509142053445 name: Cosine Precision@3 - type: cosine_precision@5 value: 0.07780590717299578 name: Cosine Precision@5 - type: cosine_precision@10 value: 0.05620253164556962 name: Cosine Precision@10 - type: cosine_recall@1 value: 0.14177215189873418 name: Cosine Recall@1 - type: cosine_recall@3 value: 0.28375527426160335 name: Cosine Recall@3 - type: cosine_recall@5 value: 0.3890295358649789 name: Cosine Recall@5 - type: cosine_recall@10 value: 0.5620253164556962 name: Cosine Recall@10 - type: cosine_ndcg@10 value: 0.322564230377663 name: Cosine Ndcg@10 - type: cosine_mrr@10 value: 0.24968421405130298 name: Cosine Mrr@10 - type: cosine_map@100 value: 0.26885741426647297 name: Cosine Map@100 - task: type: information-retrieval name: Information Retrieval dataset: name: dim 256 type: dim_256 metrics: - type: cosine_accuracy@1 value: 0.14345991561181434 name: Cosine Accuracy@1 - type: cosine_accuracy@3 value: 0.2831223628691983 name: Cosine Accuracy@3 - type: cosine_accuracy@5 value: 0.3850210970464135 name: Cosine Accuracy@5 - type: cosine_accuracy@10 value: 0.5550632911392405 name: Cosine Accuracy@10 - type: cosine_precision@1 value: 0.14345991561181434 name: Cosine Precision@1 - type: cosine_precision@3 value: 0.09437412095639944 name: Cosine Precision@3 - type: cosine_precision@5 value: 0.0770042194092827 name: Cosine Precision@5 - type: cosine_precision@10 value: 0.05550632911392406 name: Cosine Precision@10 - type: cosine_recall@1 value: 0.14345991561181434 name: Cosine Recall@1 - type: cosine_recall@3 value: 0.2831223628691983 name: Cosine Recall@3 - type: cosine_recall@5 value: 0.3850210970464135 name: Cosine Recall@5 - type: cosine_recall@10 value: 0.5550632911392405 name: Cosine Recall@10 - type: cosine_ndcg@10 value: 0.3205268083804564 name: Cosine Ndcg@10 - type: cosine_mrr@10 value: 0.24917821981113142 name: Cosine Mrr@10 - type: cosine_map@100 value: 0.2685327848764784 name: Cosine Map@100 - task: type: information-retrieval name: Information Retrieval dataset: name: dim 128 type: dim_128 metrics: - type: cosine_accuracy@1 value: 0.13924050632911392 name: Cosine Accuracy@1 - type: cosine_accuracy@3 value: 0.2795358649789029 name: Cosine Accuracy@3 - type: cosine_accuracy@5 value: 0.3837552742616034 name: Cosine Accuracy@5 - type: cosine_accuracy@10 value: 0.5533755274261604 name: Cosine Accuracy@10 - type: cosine_precision@1 value: 0.13924050632911392 name: Cosine Precision@1 - type: cosine_precision@3 value: 0.09317862165963431 name: Cosine Precision@3 - type: cosine_precision@5 value: 0.07675105485232067 name: Cosine Precision@5 - type: cosine_precision@10 value: 0.05533755274261602 name: Cosine Precision@10 - type: cosine_recall@1 value: 0.13924050632911392 name: Cosine Recall@1 - type: cosine_recall@3 value: 0.2795358649789029 name: Cosine Recall@3 - type: cosine_recall@5 value: 0.3837552742616034 name: Cosine Recall@5 - type: cosine_recall@10 value: 0.5533755274261604 name: Cosine Recall@10 - type: cosine_ndcg@10 value: 0.31759054947613424 name: Cosine Ndcg@10 - type: cosine_mrr@10 value: 0.2457681166700155 name: Cosine Mrr@10 - type: cosine_map@100 value: 0.2649300065982546 name: Cosine Map@100 - task: type: information-retrieval name: Information Retrieval dataset: name: dim 64 type: dim_64 metrics: - type: cosine_accuracy@1 value: 0.14029535864978904 name: Cosine Accuracy@1 - type: cosine_accuracy@3 value: 0.27531645569620256 name: Cosine Accuracy@3 - type: cosine_accuracy@5 value: 0.369831223628692 name: Cosine Accuracy@5 - type: cosine_accuracy@10 value: 0.5360759493670886 name: Cosine Accuracy@10 - type: cosine_precision@1 value: 0.14029535864978904 name: Cosine Precision@1 - type: cosine_precision@3 value: 0.09177215189873417 name: Cosine Precision@3 - type: cosine_precision@5 value: 0.0739662447257384 name: Cosine Precision@5 - type: cosine_precision@10 value: 0.053607594936708865 name: Cosine Precision@10 - type: cosine_recall@1 value: 0.14029535864978904 name: Cosine Recall@1 - type: cosine_recall@3 value: 0.27531645569620256 name: Cosine Recall@3 - type: cosine_recall@5 value: 0.369831223628692 name: Cosine Recall@5 - type: cosine_recall@10 value: 0.5360759493670886 name: Cosine Recall@10 - type: cosine_ndcg@10 value: 0.3099216271465372 name: Cosine Ndcg@10 - type: cosine_mrr@10 value: 0.24117783470631593 name: Cosine Mrr@10 - type: cosine_map@100 value: 0.2601649646918979 name: Cosine Map@100 --- # SentenceTransformer based on BAAI/bge-m3 This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [BAAI/bge-m3](https://huggingface.co/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](https://huggingface.co/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](https://sbert.net) - **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers) - **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers) ### 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: ```bash pip install -U sentence-transformers ``` Then you can load this model and run inference. ```python from sentence_transformers import SentenceTransformer # Download from the 🤗 Hub model = SentenceTransformer("adriansanz/ST-tramits-sitges-005-5ep") # Run inference sentences = [ 'Les queixes, observacions i suggeriments són una eina important per a millorar la qualitat dels serveis municipals.', 'Què és el que es busca amb les queixes, observacions i suggeriments?', 'Quin és el propòsit dels ajuts econòmics?', ] 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](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator) | Metric | Value | |:--------------------|:-----------| | cosine_accuracy@1 | 0.1437 | | cosine_accuracy@3 | 0.2819 | | cosine_accuracy@5 | 0.393 | | cosine_accuracy@10 | 0.5665 | | cosine_precision@1 | 0.1437 | | cosine_precision@3 | 0.094 | | cosine_precision@5 | 0.0786 | | cosine_precision@10 | 0.0566 | | cosine_recall@1 | 0.1437 | | cosine_recall@3 | 0.2819 | | cosine_recall@5 | 0.393 | | cosine_recall@10 | 0.5665 | | cosine_ndcg@10 | 0.3243 | | cosine_mrr@10 | 0.2507 | | **cosine_map@100** | **0.2695** | #### Information Retrieval * Dataset: `dim_768` * Evaluated with [InformationRetrievalEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator) | Metric | Value | |:--------------------|:-----------| | cosine_accuracy@1 | 0.147 | | cosine_accuracy@3 | 0.2871 | | cosine_accuracy@5 | 0.3901 | | cosine_accuracy@10 | 0.5631 | | cosine_precision@1 | 0.147 | | cosine_precision@3 | 0.0957 | | cosine_precision@5 | 0.078 | | cosine_precision@10 | 0.0563 | | cosine_recall@1 | 0.147 | | cosine_recall@3 | 0.2871 | | cosine_recall@5 | 0.3901 | | cosine_recall@10 | 0.5631 | | cosine_ndcg@10 | 0.3255 | | cosine_mrr@10 | 0.2533 | | **cosine_map@100** | **0.2723** | #### Information Retrieval * Dataset: `dim_512` * Evaluated with [InformationRetrievalEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator) | Metric | Value | |:--------------------|:-----------| | cosine_accuracy@1 | 0.1418 | | cosine_accuracy@3 | 0.2838 | | cosine_accuracy@5 | 0.389 | | cosine_accuracy@10 | 0.562 | | cosine_precision@1 | 0.1418 | | cosine_precision@3 | 0.0946 | | cosine_precision@5 | 0.0778 | | cosine_precision@10 | 0.0562 | | cosine_recall@1 | 0.1418 | | cosine_recall@3 | 0.2838 | | cosine_recall@5 | 0.389 | | cosine_recall@10 | 0.562 | | cosine_ndcg@10 | 0.3226 | | cosine_mrr@10 | 0.2497 | | **cosine_map@100** | **0.2689** | #### Information Retrieval * Dataset: `dim_256` * Evaluated with [InformationRetrievalEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator) | Metric | Value | |:--------------------|:-----------| | cosine_accuracy@1 | 0.1435 | | cosine_accuracy@3 | 0.2831 | | cosine_accuracy@5 | 0.385 | | cosine_accuracy@10 | 0.5551 | | cosine_precision@1 | 0.1435 | | cosine_precision@3 | 0.0944 | | cosine_precision@5 | 0.077 | | cosine_precision@10 | 0.0555 | | cosine_recall@1 | 0.1435 | | cosine_recall@3 | 0.2831 | | cosine_recall@5 | 0.385 | | cosine_recall@10 | 0.5551 | | cosine_ndcg@10 | 0.3205 | | cosine_mrr@10 | 0.2492 | | **cosine_map@100** | **0.2685** | #### Information Retrieval * Dataset: `dim_128` * Evaluated with [InformationRetrievalEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator) | Metric | Value | |:--------------------|:-----------| | cosine_accuracy@1 | 0.1392 | | cosine_accuracy@3 | 0.2795 | | cosine_accuracy@5 | 0.3838 | | cosine_accuracy@10 | 0.5534 | | cosine_precision@1 | 0.1392 | | cosine_precision@3 | 0.0932 | | cosine_precision@5 | 0.0768 | | cosine_precision@10 | 0.0553 | | cosine_recall@1 | 0.1392 | | cosine_recall@3 | 0.2795 | | cosine_recall@5 | 0.3838 | | cosine_recall@10 | 0.5534 | | cosine_ndcg@10 | 0.3176 | | cosine_mrr@10 | 0.2458 | | **cosine_map@100** | **0.2649** | #### Information Retrieval * Dataset: `dim_64` * Evaluated with [InformationRetrievalEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator) | Metric | Value | |:--------------------|:-----------| | cosine_accuracy@1 | 0.1403 | | cosine_accuracy@3 | 0.2753 | | cosine_accuracy@5 | 0.3698 | | cosine_accuracy@10 | 0.5361 | | cosine_precision@1 | 0.1403 | | cosine_precision@3 | 0.0918 | | cosine_precision@5 | 0.074 | | cosine_precision@10 | 0.0536 | | cosine_recall@1 | 0.1403 | | cosine_recall@3 | 0.2753 | | cosine_recall@5 | 0.3698 | | cosine_recall@10 | 0.5361 | | cosine_ndcg@10 | 0.3099 | | cosine_mrr@10 | 0.2412 | | **cosine_map@100** | **0.2602** | ## Training Details ### Training Dataset #### json * Dataset: json * Size: 5,214 training samples * Columns: positive and anchor * Approximate statistics based on the first 1000 samples: | | positive | anchor | |:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------| | type | string | string | | details | | | * Samples: | positive | anchor | |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------| | Ajuts per la reactivació de petites empreses i persones autònomes donades d’alta al règim especial de treballadors autònoms (RETA) amb una antiguitat superior als cinc anys (COVID19) | Quin és el requisit per a les petites empreses per rebre ajuts? | | En cas de no poder desenvolupar el projecte o activitat per la qual s'ha sol·licitat la subvenció, l'entitat beneficiària pot renunciar a la subvenció. | Puc renunciar a una subvenció si ja l'he rebut? | | L’Espai Jove de Sitges és l'equipament municipal on els joves poden dur a terme iniciatives pròpies i on també es desenvolupen d’altres impulsades per la regidoria de Joventut. | Quin és el paper de la regidoria de Joventut a l'Espai Jove de Sitges? | * Loss: [MatryoshkaLoss](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#matryoshkaloss) with these parameters: ```json { "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 | |:----------:|:------:|:-------------:|:-----------------------:|:----------------------:|:----------------------:|:----------------------:|:---------------------:|:----------------------:| | 0.4908 | 10 | 3.3699 | - | - | - | - | - | - | | 0.9816 | 20 | 1.8761 | 0.2565 | 0.2430 | 0.2509 | 0.2499 | 0.2301 | 0.2567 | | 1.4724 | 30 | 1.3111 | - | - | - | - | - | - | | 1.9632 | 40 | 0.8122 | 0.2636 | 0.2578 | 0.2629 | 0.2639 | 0.2486 | 0.2654 | | 2.4540 | 50 | 0.5903 | - | - | - | - | - | - | | 2.9448 | 60 | 0.4306 | - | - | - | - | - | - | | **2.9939** | **61** | **-** | **0.2661** | **0.2636** | **0.2648** | **0.2659** | **0.2544** | **0.2694** | | 3.4356 | 70 | 0.3553 | - | - | - | - | - | - | | 3.9264 | 80 | 0.2925 | - | - | - | - | - | - | | 3.9755 | 81 | - | 0.2701 | 0.2621 | 0.2663 | 0.2706 | 0.2602 | 0.2709 | | 4.4172 | 90 | 0.2797 | - | - | - | - | - | - | | 4.9080 | 100 | 0.267 | 0.2695 | 0.2649 | 0.2685 | 0.2689 | 0.2602 | 0.2723 | * 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 ```bibtex @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 ```bibtex @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 ```bibtex @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} } ```