--- base_model: intfloat/multilingual-e5-small library_name: sentence-transformers metrics: - cosine_accuracy - cosine_accuracy_threshold - cosine_f1 - cosine_f1_threshold - cosine_precision - cosine_recall - cosine_ap - dot_accuracy - dot_accuracy_threshold - dot_f1 - dot_f1_threshold - dot_precision - dot_recall - dot_ap - manhattan_accuracy - manhattan_accuracy_threshold - manhattan_f1 - manhattan_f1_threshold - manhattan_precision - manhattan_recall - manhattan_ap - euclidean_accuracy - euclidean_accuracy_threshold - euclidean_f1 - euclidean_f1_threshold - euclidean_precision - euclidean_recall - euclidean_ap - max_accuracy - max_accuracy_threshold - max_f1 - max_f1_threshold - max_precision - max_recall - max_ap pipeline_tag: sentence-similarity tags: - sentence-transformers - sentence-similarity - feature-extraction - generated_from_trainer - dataset_size:2871 - loss:OnlineContrastiveLoss widget: - source_sentence: Stages of photosynthesis sentences: - The function helps preprocess your entire dataset at once. - You can create an index for your dataset by using [Dataset.add_faiss_index()](/docs/datasets/v2.10.0/en/package_reference/main_classes#datasets.Dataset.add_faiss_index) or [Dataset.add_elasticsearch_index()](/docs/datasets/v2.10.0/en/package_reference/main_classes#datasets.Dataset.add_elasticsearch_index) depending on the system you want to use. - What is photosynthesis? - source_sentence: Steps to erase internet history sentences: - How do I delete my browsing history? - Yes, there is a reference section available in 🤗 Datasets documentation. It covers main classes, builder classes, loading methods, table classes, logging methods, and task templates. - What is the tallest building in New York City? - source_sentence: The `StreamingDownloadManager` class is a download manager that employs the "::" separator to traverse (possibly remote) compressed files. sentences: - What is the role of a business plan in entrepreneurship? - The Hugging Face datasets library's default handler can be disabled to prevent double logging by calling the `datasets.utils.logging.enable_propagation()` function. - The `StreamingDownloadManager` class is a download manager that uses the ”::” separator to navigate through (possibly remote) compressed archives. - source_sentence: Using torch.utils.data.DataLoader, you can package the dataset and craft a collate function to group the samples into batches. sentences: - Why does understanding death philosophical? - The `_generate_examples` method is used to access and yield TAR files sequentially, and to associate the metadata in `metadata_path` with the audio files in the TAR file. - You can wrap the dataset in DataLoader using torch.utils.data.DataLoader and create a collate function to collate the samples into batches. - source_sentence: Top literature about World War II sentences: - What is the price of an iPhone 12? - Best books on World War II - When was the Declaration of Independence signed? model-index: - name: SentenceTransformer based on intfloat/multilingual-e5-small results: - task: type: binary-classification name: Binary Classification dataset: name: pair class dev type: pair-class-dev metrics: - type: cosine_accuracy value: 0.9 name: Cosine Accuracy - type: cosine_accuracy_threshold value: 0.784720778465271 name: Cosine Accuracy Threshold - type: cosine_f1 value: 0.926605504587156 name: Cosine F1 - type: cosine_f1_threshold value: 0.784720778465271 name: Cosine F1 Threshold - type: cosine_precision value: 0.8938053097345132 name: Cosine Precision - type: cosine_recall value: 0.9619047619047619 name: Cosine Recall - type: cosine_ap value: 0.9548853455786228 name: Cosine Ap - type: dot_accuracy value: 0.9 name: Dot Accuracy - type: dot_accuracy_threshold value: 0.784720778465271 name: Dot Accuracy Threshold - type: dot_f1 value: 0.926605504587156 name: Dot F1 - type: dot_f1_threshold value: 0.784720778465271 name: Dot F1 Threshold - type: dot_precision value: 0.8938053097345132 name: Dot Precision - type: dot_recall value: 0.9619047619047619 name: Dot Recall - type: dot_ap value: 0.9548853455786228 name: Dot Ap - type: manhattan_accuracy value: 0.896875 name: Manhattan Accuracy - type: manhattan_accuracy_threshold value: 9.908977508544922 name: Manhattan Accuracy Threshold - type: manhattan_f1 value: 0.9241379310344828 name: Manhattan F1 - type: manhattan_f1_threshold value: 10.13671588897705 name: Manhattan F1 Threshold - type: manhattan_precision value: 0.8933333333333333 name: Manhattan Precision - type: manhattan_recall value: 0.9571428571428572 name: Manhattan Recall - type: manhattan_ap value: 0.9549673053310541 name: Manhattan Ap - type: euclidean_accuracy value: 0.9 name: Euclidean Accuracy - type: euclidean_accuracy_threshold value: 0.6561694145202637 name: Euclidean Accuracy Threshold - type: euclidean_f1 value: 0.926605504587156 name: Euclidean F1 - type: euclidean_f1_threshold value: 0.6561694145202637 name: Euclidean F1 Threshold - type: euclidean_precision value: 0.8938053097345132 name: Euclidean Precision - type: euclidean_recall value: 0.9619047619047619 name: Euclidean Recall - type: euclidean_ap value: 0.9548853455786228 name: Euclidean Ap - type: max_accuracy value: 0.9 name: Max Accuracy - type: max_accuracy_threshold value: 9.908977508544922 name: Max Accuracy Threshold - type: max_f1 value: 0.926605504587156 name: Max F1 - type: max_f1_threshold value: 10.13671588897705 name: Max F1 Threshold - type: max_precision value: 0.8938053097345132 name: Max Precision - type: max_recall value: 0.9619047619047619 name: Max Recall - type: max_ap value: 0.9549673053310541 name: Max Ap - task: type: binary-classification name: Binary Classification dataset: name: pair class test type: pair-class-test metrics: - type: cosine_accuracy value: 0.90625 name: Cosine Accuracy - type: cosine_accuracy_threshold value: 0.8142284154891968 name: Cosine Accuracy Threshold - type: cosine_f1 value: 0.929245283018868 name: Cosine F1 - type: cosine_f1_threshold value: 0.8142284154891968 name: Cosine F1 Threshold - type: cosine_precision value: 0.9205607476635514 name: Cosine Precision - type: cosine_recall value: 0.9380952380952381 name: Cosine Recall - type: cosine_ap value: 0.9556341092519267 name: Cosine Ap - type: dot_accuracy value: 0.90625 name: Dot Accuracy - type: dot_accuracy_threshold value: 0.8142284750938416 name: Dot Accuracy Threshold - type: dot_f1 value: 0.929245283018868 name: Dot F1 - type: dot_f1_threshold value: 0.8142284750938416 name: Dot F1 Threshold - type: dot_precision value: 0.9205607476635514 name: Dot Precision - type: dot_recall value: 0.9380952380952381 name: Dot Recall - type: dot_ap value: 0.9556341092519267 name: Dot Ap - type: manhattan_accuracy value: 0.903125 name: Manhattan Accuracy - type: manhattan_accuracy_threshold value: 9.576812744140625 name: Manhattan Accuracy Threshold - type: manhattan_f1 value: 0.9270588235294117 name: Manhattan F1 - type: manhattan_f1_threshold value: 9.576812744140625 name: Manhattan F1 Threshold - type: manhattan_precision value: 0.9162790697674419 name: Manhattan Precision - type: manhattan_recall value: 0.9380952380952381 name: Manhattan Recall - type: manhattan_ap value: 0.9557652464010216 name: Manhattan Ap - type: euclidean_accuracy value: 0.90625 name: Euclidean Accuracy - type: euclidean_accuracy_threshold value: 0.609528124332428 name: Euclidean Accuracy Threshold - type: euclidean_f1 value: 0.929245283018868 name: Euclidean F1 - type: euclidean_f1_threshold value: 0.609528124332428 name: Euclidean F1 Threshold - type: euclidean_precision value: 0.9205607476635514 name: Euclidean Precision - type: euclidean_recall value: 0.9380952380952381 name: Euclidean Recall - type: euclidean_ap value: 0.9556341092519267 name: Euclidean Ap - type: max_accuracy value: 0.90625 name: Max Accuracy - type: max_accuracy_threshold value: 9.576812744140625 name: Max Accuracy Threshold - type: max_f1 value: 0.929245283018868 name: Max F1 - type: max_f1_threshold value: 9.576812744140625 name: Max F1 Threshold - type: max_precision value: 0.9205607476635514 name: Max Precision - type: max_recall value: 0.9380952380952381 name: Max Recall - type: max_ap value: 0.9557652464010216 name: Max Ap --- # SentenceTransformer based on intfloat/multilingual-e5-small This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [intfloat/multilingual-e5-small](https://huggingface.co/intfloat/multilingual-e5-small). 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:** [intfloat/multilingual-e5-small](https://huggingface.co/intfloat/multilingual-e5-small) - **Maximum Sequence Length:** 512 tokens - **Output Dimensionality:** 384 tokens - **Similarity Function:** Cosine Similarity ### 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': 512, '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}) (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("srikarvar/fine_tuned_model_7") # Run inference sentences = [ 'Top literature about World War II', 'Best books on World War II', 'What is the price of an iPhone 12?', ] 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 * Dataset: `pair-class-dev` * Evaluated with [BinaryClassificationEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator) | Metric | Value | |:-----------------------------|:----------| | cosine_accuracy | 0.9 | | cosine_accuracy_threshold | 0.7847 | | cosine_f1 | 0.9266 | | cosine_f1_threshold | 0.7847 | | cosine_precision | 0.8938 | | cosine_recall | 0.9619 | | cosine_ap | 0.9549 | | dot_accuracy | 0.9 | | dot_accuracy_threshold | 0.7847 | | dot_f1 | 0.9266 | | dot_f1_threshold | 0.7847 | | dot_precision | 0.8938 | | dot_recall | 0.9619 | | dot_ap | 0.9549 | | manhattan_accuracy | 0.8969 | | manhattan_accuracy_threshold | 9.909 | | manhattan_f1 | 0.9241 | | manhattan_f1_threshold | 10.1367 | | manhattan_precision | 0.8933 | | manhattan_recall | 0.9571 | | manhattan_ap | 0.955 | | euclidean_accuracy | 0.9 | | euclidean_accuracy_threshold | 0.6562 | | euclidean_f1 | 0.9266 | | euclidean_f1_threshold | 0.6562 | | euclidean_precision | 0.8938 | | euclidean_recall | 0.9619 | | euclidean_ap | 0.9549 | | max_accuracy | 0.9 | | max_accuracy_threshold | 9.909 | | max_f1 | 0.9266 | | max_f1_threshold | 10.1367 | | max_precision | 0.8938 | | max_recall | 0.9619 | | **max_ap** | **0.955** | #### Binary Classification * Dataset: `pair-class-test` * Evaluated with [BinaryClassificationEvaluator](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator) | Metric | Value | |:-----------------------------|:-----------| | cosine_accuracy | 0.9062 | | cosine_accuracy_threshold | 0.8142 | | cosine_f1 | 0.9292 | | cosine_f1_threshold | 0.8142 | | cosine_precision | 0.9206 | | cosine_recall | 0.9381 | | cosine_ap | 0.9556 | | dot_accuracy | 0.9062 | | dot_accuracy_threshold | 0.8142 | | dot_f1 | 0.9292 | | dot_f1_threshold | 0.8142 | | dot_precision | 0.9206 | | dot_recall | 0.9381 | | dot_ap | 0.9556 | | manhattan_accuracy | 0.9031 | | manhattan_accuracy_threshold | 9.5768 | | manhattan_f1 | 0.9271 | | manhattan_f1_threshold | 9.5768 | | manhattan_precision | 0.9163 | | manhattan_recall | 0.9381 | | manhattan_ap | 0.9558 | | euclidean_accuracy | 0.9062 | | euclidean_accuracy_threshold | 0.6095 | | euclidean_f1 | 0.9292 | | euclidean_f1_threshold | 0.6095 | | euclidean_precision | 0.9206 | | euclidean_recall | 0.9381 | | euclidean_ap | 0.9556 | | max_accuracy | 0.9062 | | max_accuracy_threshold | 9.5768 | | max_f1 | 0.9292 | | max_f1_threshold | 9.5768 | | max_precision | 0.9206 | | max_recall | 0.9381 | | **max_ap** | **0.9558** | ## Training Details ### Training Dataset #### Unnamed Dataset * Size: 2,871 training samples * Columns: sentence2, sentence1, and label * Approximate statistics based on the first 1000 samples: | | sentence2 | sentence1 | label | |:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:------------------------------------------------| | type | string | string | int | | details | | | | * Samples: | sentence2 | sentence1 | label | |:------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------| | How do I do to get fuller face? | How can one get a fuller face? | 1 | | The DatasetInfo holds the data of a dataset, which may include its description, characteristics, and size. | A dataset's information is stored inside DatasetInfo and can include information such as the dataset description, features, and dataset size. | 1 | | How do I write a resume? | How do I create a resume? | 1 | * Loss: [OnlineContrastiveLoss](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#onlinecontrastiveloss) ### Evaluation Dataset #### Unnamed Dataset * Size: 320 evaluation samples * Columns: sentence2, sentence1, and label * Approximate statistics based on the first 320 samples: | | sentence2 | sentence1 | label | |:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:------------------------------------------------| | type | string | string | int | | details | | | | * Samples: | sentence2 | sentence1 | label | |:------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------|:---------------| | Steps to erase internet history | How do I delete my browsing history? | 1 | | How important is it to be the first person to wish someone a happy birthday? | What is the right etiquette for wishing a Jehovah Witness happy birthday? | 0 | | Who directed 'Gone with the Wind'? | Who directed 'Citizen Kane'? | 0 | * Loss: [OnlineContrastiveLoss](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#onlinecontrastiveloss) ### Training Hyperparameters #### Non-Default Hyperparameters - `eval_strategy`: epoch - `per_device_train_batch_size`: 32 - `per_device_eval_batch_size`: 32 - `gradient_accumulation_steps`: 2 - `num_train_epochs`: 4 - `warmup_ratio`: 0.1 - `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`: 32 - `per_device_eval_batch_size`: 32 - `per_gpu_train_batch_size`: None - `per_gpu_eval_batch_size`: None - `gradient_accumulation_steps`: 2 - `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`: 4 - `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`: False - `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`: 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 | loss | pair-class-dev_max_ap | pair-class-test_max_ap | |:-------:|:------:|:-------------:|:----------:|:---------------------:|:----------------------:| | 0 | 0 | - | - | 0.8735 | - | | 0.2222 | 10 | 1.3298 | - | - | - | | 0.4444 | 20 | 0.8218 | - | - | - | | 0.6667 | 30 | 0.642 | - | - | - | | 0.8889 | 40 | 0.571 | - | - | - | | 1.0 | 45 | - | 0.5321 | 0.9499 | - | | 1.1111 | 50 | 0.4828 | - | - | - | | 1.3333 | 60 | 0.3003 | - | - | - | | 1.5556 | 70 | 0.3331 | - | - | - | | 1.7778 | 80 | 0.203 | - | - | - | | **2.0** | **90** | **0.3539** | **0.5118** | **0.9558** | **-** | | 2.2222 | 100 | 0.1357 | - | - | - | | 2.4444 | 110 | 0.1562 | - | - | - | | 2.6667 | 120 | 0.0703 | - | - | - | | 2.8889 | 130 | 0.0806 | - | - | - | | 3.0 | 135 | - | 0.5266 | 0.9548 | - | | 3.1111 | 140 | 0.1721 | - | - | - | | 3.3333 | 150 | 0.1063 | - | - | - | | 3.5556 | 160 | 0.0909 | - | - | - | | 3.7778 | 170 | 0.0358 | - | - | - | | 4.0 | 180 | 0.1021 | 0.5256 | 0.9550 | 0.9558 | * The bold row denotes the saved checkpoint. ### Framework Versions - Python: 3.10.12 - Sentence Transformers: 3.1.0 - Transformers: 4.41.2 - PyTorch: 2.1.2+cu121 - Accelerate: 0.34.2 - Datasets: 2.19.1 - 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", } ```