--- tags: - sentence-transformers - sentence-similarity - feature-extraction - dense - generated_from_trainer - dataset_size:76932 - loss:MultipleNegativesRankingLoss base_model: intfloat/multilingual-e5-large widget: - source_sentence: 'query: ATM Adaptation Layer 2의 약어는 무엇인가요?' sentences: - 'passage: 2 Transmit 2 Receive (기술)' - 'passage: Alternating Current (개념)' - 'passage: AAL2 (기술)' - source_sentence: 'query: AC의 접근 클래스 C0부터 C15까지의 기능은 무엇인가요?' sentences: - 'passage: Access Class (C0 to C15) (개념)' - 'passage: 3 Dimension-Through Silicon Via (기술)' - 'passage: ACAP (Conceptual)' - source_sentence: 'query: What is the abbreviation for Alarm Agent Handling Block?' sentences: - 'passage: ATM Connection establishment/release Control Block (기술)' - 'passage: AAGHB (Technical)' - 'passage: Account Card Calling (활용)' - source_sentence: 'query: ABPL의 ATM 기본 속도 물리 계층 장치는 어떻게 구성되어 있나요?' sentences: - 'passage: ATM Base Rate Physical Layer Unit (기술)' - 'passage: 3A (개념)' - 'passage: 5GTF (Conceptual)' - source_sentence: 'query: How does the triple encryption process of 3-DES enhance security?' sentences: - 'passage: 5th Generation Technical Forum (Conceptual)' - 'passage: Triple Data Encryption Standard (Technical)' - 'passage: ABCDEF (활용)' pipeline_tag: sentence-similarity 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 model-index: - name: SentenceTransformer based on intfloat/multilingual-e5-large results: - task: type: information-retrieval name: Information Retrieval dataset: name: e5 eval real type: e5-eval-real metrics: - type: cosine_accuracy@1 value: 0.8686666666666667 name: Cosine Accuracy@1 - type: cosine_accuracy@3 value: 0.969 name: Cosine Accuracy@3 - type: cosine_accuracy@5 value: 0.9832 name: Cosine Accuracy@5 - type: cosine_accuracy@10 value: 0.9922 name: Cosine Accuracy@10 - type: cosine_precision@1 value: 0.8686666666666667 name: Cosine Precision@1 - type: cosine_precision@3 value: 0.323 name: Cosine Precision@3 - type: cosine_precision@5 value: 0.19664000000000004 name: Cosine Precision@5 - type: cosine_precision@10 value: 0.09922000000000002 name: Cosine Precision@10 - type: cosine_recall@1 value: 0.8686666666666667 name: Cosine Recall@1 - type: cosine_recall@3 value: 0.969 name: Cosine Recall@3 - type: cosine_recall@5 value: 0.9832 name: Cosine Recall@5 - type: cosine_recall@10 value: 0.9922 name: Cosine Recall@10 - type: cosine_ndcg@10 value: 0.9376619313817377 name: Cosine Ndcg@10 - type: cosine_mrr@10 value: 0.9193550000000039 name: Cosine Mrr@10 - type: cosine_map@100 value: 0.9197550584627825 name: Cosine Map@100 --- # SentenceTransformer based on intfloat/multilingual-e5-large This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [intfloat/multilingual-e5-large](https://huggingface.co/intfloat/multilingual-e5-large) on the train 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:** [intfloat/multilingual-e5-large](https://huggingface.co/intfloat/multilingual-e5-large) - **Maximum Sequence Length:** 256 tokens - **Output Dimensionality:** 1024 dimensions - **Similarity Function:** Cosine Similarity - **Training Dataset:** - train ### 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': 256, 'do_lower_case': False, 'architecture': 'XLMRobertaModel'}) (1): Pooling({'word_embedding_dimension': 1024, '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}) (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("sentence_transformers_model_id") # Run inference sentences = [ 'query: How does the triple encryption process of 3-DES enhance security?', 'passage: Triple Data Encryption Standard (Technical)', 'passage: ABCDEF (활용)', ] embeddings = model.encode(sentences) print(embeddings.shape) # [3, 1024] # Get the similarity scores for the embeddings similarities = model.similarity(embeddings, embeddings) print(similarities) # tensor([[1.0000, 0.8389, 0.1546], # [0.8389, 1.0000, 0.0850], # [0.1546, 0.0850, 1.0000]]) ``` ## Evaluation ### Metrics #### Information Retrieval * Dataset: `e5-eval-real` * Evaluated with [InformationRetrievalEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator) | Metric | Value | |:--------------------|:-----------| | cosine_accuracy@1 | 0.8687 | | cosine_accuracy@3 | 0.969 | | cosine_accuracy@5 | 0.9832 | | cosine_accuracy@10 | 0.9922 | | cosine_precision@1 | 0.8687 | | cosine_precision@3 | 0.323 | | cosine_precision@5 | 0.1966 | | cosine_precision@10 | 0.0992 | | cosine_recall@1 | 0.8687 | | cosine_recall@3 | 0.969 | | cosine_recall@5 | 0.9832 | | cosine_recall@10 | 0.9922 | | **cosine_ndcg@10** | **0.9377** | | cosine_mrr@10 | 0.9194 | | cosine_map@100 | 0.9198 | ## Training Details ### Training Dataset #### train * Dataset: train * Size: 76,932 training samples * Columns: 0 and 1 * Approximate statistics based on the first 1000 samples: | | 0 | 1 | |:--------|:-----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------| | type | string | string | | details | | | * Samples: | 0 | 1 | |:--------------------------------------------------------------------|:------------------------------------------------------------------| | query: 3D-TSV 기술의 구조는 어떻게 되어 있나요? | passage: 3 Dimension-Through Silicon Via (기술) | | query: What is the structure of the 3D-TSV technology? | passage: 3 Dimension-Through Silicon Via (Technical) | | query: 3 Dimension-Through Silicon Via의 줄임말이 뭐죠? | passage: 3D-TSV (기술) | * Loss: [MultipleNegativesRankingLoss](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters: ```json { "scale": 20.0, "similarity_fct": "cos_sim", "gather_across_devices": false } ``` ### Training Hyperparameters #### Non-Default Hyperparameters - `eval_strategy`: steps - `per_device_train_batch_size`: 64 - `per_device_eval_batch_size`: 64 - `learning_rate`: 1e-05 - `weight_decay`: 0.01 - `lr_scheduler_type`: cosine - `warmup_ratio`: 0.1 - `bf16`: True - `batch_sampler`: no_duplicates #### 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`: 64 - `per_gpu_train_batch_size`: None - `per_gpu_eval_batch_size`: None - `gradient_accumulation_steps`: 1 - `eval_accumulation_steps`: None - `torch_empty_cache_steps`: None - `learning_rate`: 1e-05 - `weight_decay`: 0.01 - `adam_beta1`: 0.9 - `adam_beta2`: 0.999 - `adam_epsilon`: 1e-08 - `max_grad_norm`: 1.0 - `num_train_epochs`: 3 - `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`: 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} - `parallelism_config`: 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`: None - `hub_always_push`: False - `hub_revision`: None - `gradient_checkpointing`: False - `gradient_checkpointing_kwargs`: None - `include_inputs_for_metrics`: False - `include_for_metrics`: [] - `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 - `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 - `use_liger_kernel`: False - `liger_kernel_config`: None - `eval_use_gather_object`: False - `average_tokens_across_devices`: False - `prompts`: None - `batch_sampler`: no_duplicates - `multi_dataset_batch_sampler`: proportional - `router_mapping`: {} - `learning_rate_mapping`: {}
### Training Logs | Epoch | Step | Training Loss | e5-eval-real_cosine_ndcg@10 | |:------:|:----:|:-------------:|:---------------------------:| | 0.0008 | 1 | 3.1575 | - | | 0.0831 | 100 | 1.6593 | - | | 0.1663 | 200 | 0.1298 | 0.8389 | | 0.2494 | 300 | 0.0848 | - | | 0.3325 | 400 | 0.0716 | 0.8808 | | 0.4156 | 500 | 0.0504 | - | | 0.4988 | 600 | 0.0421 | 0.9033 | | 0.5819 | 700 | 0.042 | - | | 0.6650 | 800 | 0.0398 | 0.9095 | | 0.7481 | 900 | 0.0384 | - | | 0.8313 | 1000 | 0.0383 | 0.9111 | | 0.9144 | 1100 | 0.0321 | - | | 0.9975 | 1200 | 0.0317 | 0.9186 | | 1.0806 | 1300 | 0.0299 | - | | 1.1638 | 1400 | 0.0302 | 0.9161 | | 1.2469 | 1500 | 0.025 | - | | 1.3300 | 1600 | 0.0199 | 0.9261 | | 1.4131 | 1700 | 0.0179 | - | | 1.4963 | 1800 | 0.0117 | 0.9305 | | 1.5794 | 1900 | 0.013 | - | | 1.6625 | 2000 | 0.012 | 0.9308 | | 1.7456 | 2100 | 0.0137 | - | | 1.8288 | 2200 | 0.0141 | 0.9309 | | 1.9119 | 2300 | 0.0127 | - | | 1.9950 | 2400 | 0.0115 | 0.9332 | | 2.0781 | 2500 | 0.0114 | - | | 2.1613 | 2600 | 0.011 | 0.9351 | | 2.2444 | 2700 | 0.0107 | - | | 2.3275 | 2800 | 0.0087 | 0.9357 | | 2.4106 | 2900 | 0.0084 | - | | 2.4938 | 3000 | 0.0059 | 0.9366 | | 2.5769 | 3100 | 0.0062 | - | | 2.6600 | 3200 | 0.0071 | 0.9377 | | 2.7431 | 3300 | 0.0072 | - | | 2.8263 | 3400 | 0.0079 | 0.9376 | | 2.9094 | 3500 | 0.0071 | - | | 2.9925 | 3600 | 0.0068 | 0.9376 | | -1 | -1 | - | 0.9377 | ### Framework Versions - Python: 3.12.11 - Sentence Transformers: 5.1.0 - Transformers: 4.56.1 - PyTorch: 2.8.0+cu126 - Accelerate: 1.10.1 - Datasets: 3.6.0 - Tokenizers: 0.22.0 ## 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", } ``` #### 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} } ```