tomaarsen/mpnet-base-gooaq-cmnrl-mrl

MPNet base trained on GooAQ triplets

This is a sentence-transformers model finetuned from microsoft/mpnet-base on the gooaq dataset. It maps sentences & paragraphs to a 768-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: microsoft/mpnet-base
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • gooaq
• Language: en
• License: apache-2.0

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': 512, 'do_lower_case': False}) with Transformer model: MPNetModel 
  (1): Pooling({'word_embedding_dimension': 768, '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("tomaarsen/mpnet-base-gooaq-cmnrl-mrl")
# Run inference
sentences = [
    "how to reverse a video on tiktok that's not yours?",
    '[\'Tap "Effects" at the bottom of your screen — it\\\'s an icon that looks like a clock. Open the Effects menu. ... \', \'At the end of the new list that appears, tap "Time." Select "Time" at the end. ... \', \'Select "Reverse" — you\\\'ll then see a preview of your new, reversed video appear on the screen.\']',
    'Relative age is the age of a rock layer (or the fossils it contains) compared to other layers. It can be determined by looking at the position of rock layers. Absolute age is the numeric age of a layer of rocks or fossils. Absolute age can be determined by using radiometric dating.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

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

Evaluation

Metrics

Information Retrieval

• Datasets: NanoClimateFEVER, NanoDBPedia, NanoFEVER, NanoFiQA2018, NanoHotpotQA, NanoMSMARCO, NanoNFCorpus, NanoNQ, NanoQuoraRetrieval, NanoSCIDOCS, NanoArguAna, NanoSciFact and NanoTouche2020
• Evaluated with InformationRetrievalEvaluator

Metric	NanoClimateFEVER	NanoDBPedia	NanoFEVER	NanoFiQA2018	NanoHotpotQA	NanoMSMARCO	NanoNFCorpus	NanoNQ	NanoQuoraRetrieval	NanoSCIDOCS	NanoArguAna	NanoSciFact	NanoTouche2020
cosine_accuracy@1	0.26	0.56	0.62	0.4	0.6	0.26	0.42	0.46	0.92	0.34	0.24	0.52	0.5102
cosine_accuracy@3	0.46	0.78	0.82	0.52	0.72	0.54	0.52	0.64	0.98	0.54	0.5	0.64	0.8163
cosine_accuracy@5	0.5	0.82	0.84	0.6	0.78	0.7	0.54	0.68	0.98	0.64	0.68	0.72	0.8571
cosine_accuracy@10	0.62	0.88	0.9	0.68	0.84	0.82	0.64	0.8	1.0	0.76	0.82	0.74	0.9796
cosine_precision@1	0.26	0.56	0.62	0.4	0.6	0.26	0.42	0.46	0.92	0.34	0.24	0.52	0.5102
cosine_precision@3	0.1733	0.5	0.28	0.26	0.32	0.18	0.3533	0.2267	0.4067	0.26	0.1667	0.2267	0.5102
cosine_precision@5	0.116	0.436	0.172	0.188	0.204	0.14	0.296	0.144	0.256	0.216	0.136	0.16	0.4735
cosine_precision@10	0.082	0.378	0.092	0.112	0.118	0.082	0.23	0.084	0.134	0.148	0.082	0.084	0.4102
cosine_recall@1	0.1283	0.0541	0.5767	0.2439	0.3	0.26	0.0248	0.44	0.7973	0.07	0.24	0.485	0.0389
cosine_recall@3	0.2357	0.1204	0.7867	0.3761	0.48	0.54	0.0501	0.63	0.9453	0.16	0.5	0.61	0.1159
cosine_recall@5	0.2523	0.1593	0.8067	0.4295	0.51	0.7	0.0635	0.67	0.9593	0.2227	0.68	0.705	0.1756
cosine_recall@10	0.3423	0.237	0.8667	0.5026	0.59	0.82	0.0885	0.76	0.9893	0.3047	0.82	0.73	0.2858
cosine_ndcg@10	0.2832	0.4606	0.7422	0.4396	0.5464	0.5254	0.2784	0.6103	0.9468	0.2918	0.5108	0.6182	0.4589
cosine_mrr@10	0.3686	0.6702	0.7256	0.4848	0.6749	0.4324	0.482	0.5662	0.9489	0.4679	0.4136	0.5913	0.6822
cosine_map@100	0.2282	0.3135	0.6985	0.3961	0.4778	0.4419	0.1099	0.5687	0.9245	0.2098	0.4235	0.5834	0.3613

Nano BEIR

• Dataset: NanoBEIR_mean
• Evaluated with NanoBEIREvaluator

Metric	Value
cosine_accuracy@1	0.47
cosine_accuracy@3	0.652
cosine_accuracy@5	0.7182
cosine_accuracy@10	0.8061
cosine_precision@1	0.47
cosine_precision@3	0.2972
cosine_precision@5	0.226
cosine_precision@10	0.1566
cosine_recall@1	0.2815
cosine_recall@3	0.4269
cosine_recall@5	0.4872
cosine_recall@10	0.5644
cosine_ndcg@10	0.5163
cosine_mrr@10	0.5776
cosine_map@100	0.4413

Training Details

Training Dataset

gooaq

• Dataset: gooaq at b089f72
• Size: 3,012,496 training samples
• Columns: question and answer
• Approximate statistics based on the first 1000 samples:

  	question	answer
  type	string	string
  details	min: 8 tokens mean: 11.86 tokens max: 21 tokens	min: 14 tokens mean: 60.48 tokens max: 138 tokens

• Samples:

  question	answer
  what is the difference between broilers and layers?	An egg laying poultry is called egger or layer whereas broilers are reared for obtaining meat. So a layer should be able to produce more number of large sized eggs, without growing too much. On the other hand, a broiler should yield more meat and hence should be able to grow well.
  what is the difference between chronological order and spatial order?	As a writer, you should always remember that unlike chronological order and the other organizational methods for data, spatial order does not take into account the time. Spatial order is primarily focused on the location. All it does is take into account the location of objects and not the time.
  is kamagra same as viagra?	Kamagra is thought to contain the same active ingredient as Viagra, sildenafil citrate. In theory, it should work in much the same way as Viagra, taking about 45 minutes to take effect, and lasting for around 4-6 hours. However, this will vary from person to person.

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "CachedMultipleNegativesRankingLoss",
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Evaluation Dataset

gooaq

• Dataset: gooaq at b089f72
• Size: 3,012,496 evaluation samples
• Columns: question and answer
• Approximate statistics based on the first 1000 samples:

  	question	answer
  type	string	string
  details	min: 8 tokens mean: 11.88 tokens max: 22 tokens	min: 14 tokens mean: 61.03 tokens max: 127 tokens

• Samples:

  question	answer
  how do i program my directv remote with my tv?	['Press MENU on your remote.', 'Select Settings & Help > Settings > Remote Control > Program Remote.', 'Choose the device (TV, audio, DVD) you wish to program. ... ', 'Follow the on-screen prompts to complete programming.']
  are rodrigues fruit bats nocturnal?	Before its numbers were threatened by habitat destruction, storms, and hunting, some of those groups could number 500 or more members. Sunrise, sunset. Rodrigues fruit bats are most active at dawn, at dusk, and at night.
  why does your heart rate increase during exercise bbc bitesize?	During exercise there is an increase in physical activity and muscle cells respire more than they do when the body is at rest. The heart rate increases during exercise. The rate and depth of breathing increases - this makes sure that more oxygen is absorbed into the blood, and more carbon dioxide is removed from it.

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "CachedMultipleNegativesRankingLoss",
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 2048
• per_device_eval_batch_size: 2048
• learning_rate: 8e-05
• num_train_epochs: 1
• 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: 2048
• per_device_eval_batch_size: 2048
• 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: 8e-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: 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
• 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: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	NanoClimateFEVER_cosine_ndcg@10	NanoDBPedia_cosine_ndcg@10	NanoFEVER_cosine_ndcg@10	NanoFiQA2018_cosine_ndcg@10	NanoHotpotQA_cosine_ndcg@10	NanoMSMARCO_cosine_ndcg@10	NanoNFCorpus_cosine_ndcg@10	NanoNQ_cosine_ndcg@10	NanoQuoraRetrieval_cosine_ndcg@10	NanoSCIDOCS_cosine_ndcg@10	NanoArguAna_cosine_ndcg@10	NanoSciFact_cosine_ndcg@10	NanoTouche2020_cosine_ndcg@10	NanoBEIR_mean_cosine_ndcg@10
0	0	-	-	0.0419	0.1123	0.0389	0.0309	0.0746	0.1310	0.0311	0.0397	0.6607	0.0638	0.2616	0.1097	0.1098	0.1312
0.0007	1	41.9671	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.0682	100	12.4237	1.0176	0.3022	0.4597	0.7934	0.4621	0.5280	0.4849	0.2517	0.5561	0.8988	0.3144	0.5708	0.5755	0.4514	0.5115
0.1363	200	3.0536	0.6917	0.2883	0.4588	0.7773	0.4272	0.5264	0.5494	0.2538	0.5837	0.9303	0.2945	0.5493	0.5795	0.4547	0.5133
0.2045	300	2.2724	0.5954	0.2944	0.4606	0.7825	0.4522	0.5247	0.5069	0.2554	0.5636	0.9177	0.2861	0.5560	0.5562	0.4667	0.5095
0.2727	400	1.933	0.5171	0.3027	0.4841	0.7050	0.4406	0.4877	0.5406	0.2768	0.6014	0.9463	0.2989	0.5725	0.6151	0.4680	0.5184
0.3408	500	1.7806	0.4745	0.3034	0.4857	0.7537	0.4435	0.5661	0.5529	0.2733	0.5878	0.9470	0.3016	0.5377	0.6073	0.4682	0.5252
0.4090	600	1.6253	0.4392	0.3018	0.4790	0.7502	0.4617	0.5478	0.5411	0.2812	0.6220	0.9443	0.2916	0.5210	0.5900	0.4644	0.5228
0.4772	700	1.5136	0.4312	0.3175	0.4846	0.7481	0.4168	0.5761	0.5222	0.2825	0.6142	0.9415	0.2888	0.5373	0.5754	0.4675	0.5210
0.5453	800	1.4454	0.4022	0.3017	0.4756	0.7307	0.4494	0.5484	0.5184	0.2821	0.6182	0.9440	0.2834	0.5191	0.6071	0.4694	0.5191
0.6135	900	1.3711	0.3886	0.2945	0.4602	0.7463	0.4529	0.5433	0.5457	0.2730	0.5972	0.9449	0.2776	0.5183	0.6018	0.4716	0.5175
0.6817	1000	1.3295	0.3688	0.2811	0.4720	0.7275	0.4342	0.5581	0.5418	0.2809	0.6087	0.9421	0.2823	0.5138	0.5729	0.4662	0.5140
0.7498	1100	1.267	0.3637	0.2815	0.4666	0.7168	0.4346	0.5348	0.5317	0.2789	0.6056	0.9450	0.2775	0.5117	0.6116	0.4583	0.5119
0.8180	1200	1.2542	0.3514	0.2882	0.4659	0.7275	0.4308	0.5585	0.5373	0.2788	0.5950	0.9433	0.2767	0.5241	0.6141	0.4655	0.5158
0.8862	1300	1.2146	0.3427	0.2932	0.4638	0.7118	0.4453	0.5636	0.5363	0.2788	0.6098	0.9481	0.2825	0.5160	0.6238	0.4619	0.5181
0.9543	1400	1.1892	0.3378	0.2809	0.4610	0.7319	0.4353	0.5397	0.5295	0.2828	0.6029	0.9474	0.2931	0.5078	0.6182	0.4602	0.5147
1.0	1467	-	-	0.2832	0.4606	0.7422	0.4396	0.5464	0.5254	0.2784	0.6103	0.9468	0.2918	0.5108	0.6182	0.4589	0.5163

Environmental Impact

Carbon emissions were measured using CodeCarbon.

• Energy Consumed: 2.318 kWh
• Carbon Emitted: 0.901 kg of CO2
• Hours Used: 5.999 hours

Training Hardware

• On Cloud: No
• GPU Model: 1 x NVIDIA GeForce RTX 3090
• CPU Model: 13th Gen Intel(R) Core(TM) i7-13700K
• RAM Size: 31.78 GB

Framework Versions

• Python: 3.11.6
• Sentence Transformers: 3.4.0.dev0
• Transformers: 4.46.2
• PyTorch: 2.5.0+cu121
• Accelerate: 1.1.1
• Datasets: 2.20.0
• Tokenizers: 0.20.3

Citation

BibTeX

Sentence Transformers

@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

@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}
}

CachedMultipleNegativesRankingLoss

@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}