--- language: - en tags: - sentence-transformers - cross-encoder - generated_from_trainer - dataset_size:78704 - loss:ListMLELoss base_model: microsoft/MiniLM-L12-H384-uncased datasets: - microsoft/ms_marco pipeline_tag: text-ranking library_name: sentence-transformers metrics: - map - mrr@10 - ndcg@10 model-index: - name: CrossEncoder based on microsoft/MiniLM-L12-H384-uncased results: - task: type: cross-encoder-reranking name: Cross Encoder Reranking dataset: name: NanoMSMARCO R100 type: NanoMSMARCO_R100 metrics: - type: map value: 0.519 name: Map - type: mrr@10 value: 0.5072 name: Mrr@10 - type: ndcg@10 value: 0.5754 name: Ndcg@10 - task: type: cross-encoder-reranking name: Cross Encoder Reranking dataset: name: NanoNFCorpus R100 type: NanoNFCorpus_R100 metrics: - type: map value: 0.3333 name: Map - type: mrr@10 value: 0.5492 name: Mrr@10 - type: ndcg@10 value: 0.353 name: Ndcg@10 - task: type: cross-encoder-reranking name: Cross Encoder Reranking dataset: name: NanoNQ R100 type: NanoNQ_R100 metrics: - type: map value: 0.5948 name: Map - type: mrr@10 value: 0.5977 name: Mrr@10 - type: ndcg@10 value: 0.6497 name: Ndcg@10 - task: type: cross-encoder-nano-beir name: Cross Encoder Nano BEIR dataset: name: NanoBEIR R100 mean type: NanoBEIR_R100_mean metrics: - type: map value: 0.4824 name: Map - type: mrr@10 value: 0.5513 name: Mrr@10 - type: ndcg@10 value: 0.526 name: Ndcg@10 --- # CrossEncoder based on microsoft/MiniLM-L12-H384-uncased This is a [Cross Encoder](https://www.sbert.net/docs/cross_encoder/usage/usage.html) model finetuned from [microsoft/MiniLM-L12-H384-uncased](https://huggingface.co/microsoft/MiniLM-L12-H384-uncased) on the [ms_marco](https://huggingface.co/datasets/microsoft/ms_marco) dataset using the [sentence-transformers](https://www.SBERT.net) library. It computes scores for pairs of texts, which can be used for text reranking and semantic search. ## Model Details ### Model Description - **Model Type:** Cross Encoder - **Base model:** [microsoft/MiniLM-L12-H384-uncased](https://huggingface.co/microsoft/MiniLM-L12-H384-uncased) - **Maximum Sequence Length:** 512 tokens - **Number of Output Labels:** 1 label - **Training Dataset:** - [ms_marco](https://huggingface.co/datasets/microsoft/ms_marco) - **Language:** en ### Model Sources - **Documentation:** [Sentence Transformers Documentation](https://sbert.net) - **Documentation:** [Cross Encoder Documentation](https://www.sbert.net/docs/cross_encoder/usage/usage.html) - **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers) - **Hugging Face:** [Cross Encoders on Hugging Face](https://huggingface.co/models?library=sentence-transformers&other=cross-encoder) ## 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 CrossEncoder # Download from the đŸ€— Hub model = CrossEncoder("yjoonjang/reranker-msmarco-v1.1-MiniLM-L12-H384-uncased-plistmle") # Get scores for pairs of texts pairs = [ ['How many calories in an egg', 'There are on average between 55 and 80 calories in an egg depending on its size.'], ['How many calories in an egg', 'Egg whites are very low in calories, have no fat, no cholesterol, and are loaded with protein.'], ['How many calories in an egg', 'Most of the calories in an egg come from the yellow yolk in the center.'], ] scores = model.predict(pairs) print(scores.shape) # (3,) # Or rank different texts based on similarity to a single text ranks = model.rank( 'How many calories in an egg', [ 'There are on average between 55 and 80 calories in an egg depending on its size.', 'Egg whites are very low in calories, have no fat, no cholesterol, and are loaded with protein.', 'Most of the calories in an egg come from the yellow yolk in the center.', ] ) # [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...] ``` ## Evaluation ### Metrics #### Cross Encoder Reranking * Datasets: `NanoMSMARCO_R100`, `NanoNFCorpus_R100` and `NanoNQ_R100` * Evaluated with [CrossEncoderRerankingEvaluator](https://sbert.net/docs/package_reference/cross_encoder/evaluation.html#sentence_transformers.cross_encoder.evaluation.CrossEncoderRerankingEvaluator) with these parameters: ```json { "at_k": 10, "always_rerank_positives": true } ``` | Metric | NanoMSMARCO_R100 | NanoNFCorpus_R100 | NanoNQ_R100 | |:------------|:---------------------|:---------------------|:---------------------| | map | 0.5190 (+0.0295) | 0.3333 (+0.0723) | 0.5948 (+0.1752) | | mrr@10 | 0.5072 (+0.0297) | 0.5492 (+0.0493) | 0.5977 (+0.1710) | | **ndcg@10** | **0.5754 (+0.0350)** | **0.3530 (+0.0280)** | **0.6497 (+0.1491)** | #### Cross Encoder Nano BEIR * Dataset: `NanoBEIR_R100_mean` * Evaluated with [CrossEncoderNanoBEIREvaluator](https://sbert.net/docs/package_reference/cross_encoder/evaluation.html#sentence_transformers.cross_encoder.evaluation.CrossEncoderNanoBEIREvaluator) with these parameters: ```json { "dataset_names": [ "msmarco", "nfcorpus", "nq" ], "rerank_k": 100, "at_k": 10, "always_rerank_positives": true } ``` | Metric | Value | |:------------|:---------------------| | map | 0.4824 (+0.0923) | | mrr@10 | 0.5513 (+0.0833) | | **ndcg@10** | **0.5260 (+0.0707)** | ## Training Details ### Training Dataset #### ms_marco * Dataset: [ms_marco](https://huggingface.co/datasets/microsoft/ms_marco) at [a47ee7a](https://huggingface.co/datasets/microsoft/ms_marco/tree/a47ee7aae8d7d466ba15f9f0bfac3b3681087b3a) * Size: 78,704 training samples * Columns: query, docs, and labels * Approximate statistics based on the first 1000 samples: | | query | docs | labels | |:--------|:------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------| | type | string | list | list | | details | | | | * Samples: | query | docs | labels | |:------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:----------------------------------| | ampullae of lorenzini definition | ['Definition of AMPULLA OF LORENZINI. : any of the pores on the snouts of marine sharks and rays that contain receptors highly sensitive to weak electric fields. ADVERTISEMENT. Stefano Lorenzini fl 1678 Italian physician. First Known Use: 1898.', 'Definition of AMPULLA. 1. : a glass or earthenware flask with a globular body and two handles used especially by the ancient Romans to hold ointment, perfume, or wine. 2. : a saccular anatomical swelling or pouch. — am·pul·la·ry \\am-ˈpu̇-lər-ē, ˈam-pyə-ˌler-ē\\ adjective.', 'These sensory organs help fish to sense electric fields in the water. Each ampulla consists of a jelly-filled canal opening to the surface by a pore in the skin and ending blindly in a cluster of small pockets full of special jelly.', 'Wiktionary (5.00 / 1 vote) Rate this definition: ampulla of Lorenzini (Noun). An electroreceptor found mainly in cartilaginous fish such as sharks and rays, forming a network of jelly-filled canals. Origin: After Stephano Lorenzini, who first described them.', 'The ampullae of Lorenzini are special sensing organs called electroreceptors, forming a network of jelly-filled pores. They are mostly discussed as being found in cartilaginous fish (sharks, rays, and chimaeras); however, they are also reported to be found in Chondrostei such as reedfish and sturgeon.'] | [1, 0, 0, 0, 0] | | pulmonary function tests are conducted by respiratory therapists | ['Respiratory Care. Our Respiratory Care Department offers a full range of inpatient therapeutic and diagnostic services, including a full range of pulmonary function testing. Our therapists also provide pulmonary education such as Living with COPD and the Asthma Awareness Program.. ', 'Spirometry. Spirometry is the first and most commonly done lung function test. It measures how much and how quickly you can move air out of your lungs. For this test, you breathe into a mouthpiece attached to a recording device (spirometer). Lung Function Tests. Guide. Lung function tests (also called pulmonary function tests, or PFTs) check how well your lungs work. The tests determine how much air your lungs can hold, how quickly you can move air in and out of your lungs, and how well your lungs put oxygen into and remove carbon dioxide from your blood.', 'They provide your physician needed information to help diagnose disease, measure the severity of lung problems, recommend treatments, and follow yo... | [1, 0, 0, 0, 0, ...] | | organization of American states definition | ["The Organization of American States, or the OAS, is a continental organization founded on 30 April 1948 for the purposes of regional solidarity and cooperation among its member states. Headquartered in Washington, D.C., United States, the OAS's members are the 35 independent states of the Americas. ", 'More videos ». The Organization of American States is the premier regional forum for political discussion, policy analysis and decision-making in Western Hemisphere affairs. The OAS brings together leaders from nations across the Americas to address hemispheric issues and opportunities. The Coordinating Office of the Offices in the Member States invites you to visit their site. You will be able to receive updates, find out who they are and learn out about projects, programs, internships, and scholarships in each office.', "That adherence by any member of the Organization of American States to Marxism-Leninism is incompatible with the inter-American system and the alignment of such a go... | [1, 0, 0, 0, 0, ...] | * Loss: [ListMLELoss](https://sbert.net/docs/package_reference/cross_encoder/losses.html#listmleloss) with these parameters: ```json { "lambda_weight": "sentence_transformers.cross_encoder.losses.ListMLELoss.ListMLELambdaWeight", "activation_fct": "torch.nn.modules.linear.Identity", "mini_batch_size": 16, "respect_input_order": true } ``` ### Evaluation Dataset #### ms_marco * Dataset: [ms_marco](https://huggingface.co/datasets/microsoft/ms_marco) at [a47ee7a](https://huggingface.co/datasets/microsoft/ms_marco/tree/a47ee7aae8d7d466ba15f9f0bfac3b3681087b3a) * Size: 1,000 evaluation samples * Columns: query, docs, and labels * Approximate statistics based on the first 1000 samples: | | query | docs | labels | |:--------|:-----------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------| | type | string | list | list | | details | | | | * Samples: | query | docs | labels | |:--------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:----------------------------------| | what is tidal flow | ['Noun. 1. tidal flow-the water current caused by the tides. tidal current. tide-the periodic rise and fall of the sea level under the gravitational pull of the moon. aegir, eager, eagre, tidal bore, bore-a high wave (often dangerous) caused by tidal flow (as by colliding tidal currents or in a narrow estuary). ', 'Tidal energy is a form of hydropower that converts the energy of the tides into electricity or other useful forms of power. The tide is created by the gravitational effect of the sun and the moon on the earth causing cyclical movement of the seas. Tidal Stream. Tidal Stream is the flow of water as the tide ebbs and floods, and manifests itself as tidal current. Tidal Stream devices seek to extract energy from this kinetic movement of water, much as wind turbines extract energy from the movement of air.', 'A horizontal movement of water often accompanies the rising and falling of the tide. This is called the tidal current. The incoming tide along the coast and into the bays a... | [1, 0, 0, 0, 0, ...] | | what is matelasse | ['The French word, matelasse matelassĂ© “means,” “quilted,” padded “or,” cushioned and in usage with, fabric refers to hand quilted. Textiles it is meant to mimic the style of-hand Stitched marseilles type quilts made In, Provence. france Matelasse matelassĂ© fabric is used on upholstery for slip covers and throw, pillows and in, bedding for, coverlets duvet covers and pillow. Shams it is also used in crib bedding and’children s bedding. sets', 'Matelasse (matelassĂ©-mat-LA) say is a weaving or stitching technique yielding a pattern that appears quilted or. Padded matelasse matelassĂ© may be achieved, by hand on a, jacquard loom or a. Quilting machine it is meant to mimic the style-of hand stitched quilts Made, In. marseilles france Matelasse matelassĂ© may be achieved by, hand on a jacquard, loom or a quilting. Machine it is meant to mimic the style of-hand stitched quilts made In, Marseilles. france', "Save. Matelasse is type of double-woven fabric that first gained popularity in the 18th... | [1, 1, 0, 0, 0, ...] | | what does atp mean | ['Conversion from ATP to ADP. Adenosine triphosphate (ATP) is the energy currency of life and it provides that energy for most biological processes by being converted to ADP (adenosine diphosphate). Since the basic reaction involves a water molecule, this reaction is commonly referred to as the hydrolysis of ATP. Free Energy from Hydrolysis of ATP. Adenosine triphosphate (ATP) is the energy currency of life and it provides that energy for most biological processes by being converted to ADP (adenosine diphosphate). Since the basic reaction involves a water molecule, this reaction is commonly referred to as the hydrolysis of ATP.', 'ATP is a nucleotide that contains a large amount of chemical energy stored in its high-energy phosphate bonds. It releases energy when it is broken down (hydrolyzed) into ADP (or Adenosine Diphosphate). The energy is used for many metabolic processes. ', '‱ ATP (noun). The noun ATP has 1 sense: 1. a nucleotide derived from adenosine that occurs in muscle tiss... | [1, 0, 0, 0, 0, ...] | * Loss: [ListMLELoss](https://sbert.net/docs/package_reference/cross_encoder/losses.html#listmleloss) with these parameters: ```json { "lambda_weight": "sentence_transformers.cross_encoder.losses.ListMLELoss.ListMLELambdaWeight", "activation_fct": "torch.nn.modules.linear.Identity", "mini_batch_size": 16, "respect_input_order": true } ``` ### Training Hyperparameters #### Non-Default Hyperparameters - `eval_strategy`: steps - `per_device_train_batch_size`: 16 - `per_device_eval_batch_size`: 16 - `learning_rate`: 2e-05 - `num_train_epochs`: 1 - `warmup_ratio`: 0.1 - `seed`: 12 - `bf16`: True - `load_best_model_at_end`: 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`: 16 - `per_device_eval_batch_size`: 16 - `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`: 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`: 1 - `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`: 12 - `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`: 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 - `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 - `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 - `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 - `use_liger_kernel`: False - `eval_use_gather_object`: False - `average_tokens_across_devices`: False - `prompts`: None - `batch_sampler`: batch_sampler - `multi_dataset_batch_sampler`: proportional
### Training Logs | Epoch | Step | Training Loss | Validation Loss | NanoMSMARCO_R100_ndcg@10 | NanoNFCorpus_R100_ndcg@10 | NanoNQ_R100_ndcg@10 | NanoBEIR_R100_mean_ndcg@10 | |:----------:|:--------:|:-------------:|:---------------:|:------------------------:|:-------------------------:|:--------------------:|:--------------------------:| | -1 | -1 | - | - | 0.0301 (-0.5103) | 0.2693 (-0.0557) | 0.0549 (-0.4457) | 0.1181 (-0.3372) | | 0.0002 | 1 | 909.2226 | - | - | - | - | - | | 0.0508 | 250 | 918.5451 | - | - | - | - | - | | 0.1016 | 500 | 883.3122 | 876.4382 | 0.2066 (-0.3338) | 0.2445 (-0.0805) | 0.3186 (-0.1821) | 0.2566 (-0.1988) | | 0.1525 | 750 | 859.0346 | - | - | - | - | - | | 0.2033 | 1000 | 864.3308 | 850.8157 | 0.4610 (-0.0794) | 0.3138 (-0.0112) | 0.6074 (+0.1068) | 0.4607 (+0.0054) | | 0.2541 | 1250 | 851.3652 | - | - | - | - | - | | 0.3049 | 1500 | 838.7614 | 838.7972 | 0.5708 (+0.0304) | 0.3423 (+0.0173) | 0.6056 (+0.1050) | 0.5063 (+0.0509) | | 0.3558 | 1750 | 853.0997 | - | - | - | - | - | | 0.4066 | 2000 | 837.1816 | 834.6595 | 0.4936 (-0.0469) | 0.3460 (+0.0209) | 0.5778 (+0.0771) | 0.4724 (+0.0171) | | 0.4574 | 2250 | 820.9718 | - | - | - | - | - | | **0.5082** | **2500** | **829.679** | **832.1774** | **0.5754 (+0.0350)** | **0.3530 (+0.0280)** | **0.6497 (+0.1491)** | **0.5260 (+0.0707)** | | 0.5591 | 2750 | 816.8598 | - | - | - | - | - | | 0.6099 | 3000 | 841.9976 | 830.9660 | 0.5351 (-0.0054) | 0.3651 (+0.0401) | 0.6357 (+0.1351) | 0.5120 (+0.0566) | | 0.6607 | 3250 | 820.7183 | - | - | - | - | - | | 0.7115 | 3500 | 812.7813 | 825.5827 | 0.5444 (+0.0040) | 0.3803 (+0.0552) | 0.6208 (+0.1201) | 0.5152 (+0.0598) | | 0.7624 | 3750 | 852.4021 | - | - | - | - | - | | 0.8132 | 4000 | 830.3532 | 824.7762 | 0.5760 (+0.0355) | 0.3600 (+0.0350) | 0.6315 (+0.1309) | 0.5225 (+0.0671) | | 0.8640 | 4250 | 834.5426 | - | - | - | - | - | | 0.9148 | 4500 | 828.2203 | 822.1611 | 0.5711 (+0.0307) | 0.3682 (+0.0432) | 0.6303 (+0.1296) | 0.5232 (+0.0678) | | 0.9656 | 4750 | 842.7682 | - | - | - | - | - | | -1 | -1 | - | - | 0.5754 (+0.0350) | 0.3530 (+0.0280) | 0.6497 (+0.1491) | 0.5260 (+0.0707) | * The bold row denotes the saved checkpoint. ### Framework Versions - Python: 3.11.11 - Sentence Transformers: 3.5.0.dev0 - Transformers: 4.49.0 - PyTorch: 2.6.0+cu124 - Accelerate: 1.5.2 - Datasets: 3.4.0 - Tokenizers: 0.21.1 ## 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", } ``` #### ListMLELoss ```bibtex @inproceedings{lan2013position, title={Position-aware ListMLE: a sequential learning process for ranking}, author={Lan, Yanyan and Guo, Jiafeng and Cheng, Xueqi and Liu, Tie-Yan}, booktitle={Proceedings of the Twenty-Ninth Conference on Uncertainty in Artificial Intelligence}, pages={333--342}, year={2013} } ```