tomaarsen/MiniLM-L6-H384-uncased-gooaq-asym

MPNet base trained on AllNLI triplets

This is a sentence-transformers model finetuned from nreimers/MiniLM-L6-H384-uncased on the gooaq dataset. 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: nreimers/MiniLM-L6-H384-uncased
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 384 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: 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})
  (asym): Asym(
    (query-0): Dense({'in_features': 384, 'out_features': 384, 'bias': True, 'activation_function': 'torch.nn.modules.activation.Tanh'})
    (doc-0): Dense({'in_features': 384, 'out_features': 384, 'bias': True, 'activation_function': 'torch.nn.modules.activation.Tanh'})
  )
)

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/MiniLM-L6-H384-uncased-gooaq-asym")
# Run inference
sentences = [
    'The weather is lovely today.',
    "It's so sunny outside!",
    'He drove to the stadium.',
]
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

Information Retrieval

• Dataset: gooaq-dev
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1588
cosine_accuracy@3	0.2785
cosine_accuracy@5	0.3457
cosine_accuracy@10	0.4466
cosine_precision@1	0.1588
cosine_precision@3	0.0928
cosine_precision@5	0.0691
cosine_precision@10	0.0447
cosine_recall@1	0.1588
cosine_recall@3	0.2785
cosine_recall@5	0.3457
cosine_recall@10	0.4466
cosine_ndcg@10	0.2882
cosine_mrr@10	0.2393
cosine_map@100	0.2521

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	dict	dict
  details

• Samples:

  question	answer
  {'query': 'what is the difference between broilers and layers?'}	{'doc': '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.'}
  {'query': 'what is the difference between chronological order and spatial order?'}	{'doc': '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.'}
  {'query': 'is kamagra same as viagra?'}	{'doc': '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: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

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	dict	dict
  details

• Samples:

  question	answer
  {'query': 'how do i program my directv remote with my tv?'}	{'doc': "['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.']"}
  {'query': 'are rodrigues fruit bats nocturnal?'}	{'doc': '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.'}
  {'query': 'why does your heart rate increase during exercise bbc bitesize?'}	{'doc': '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: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 128
• per_device_eval_batch_size: 128
• learning_rate: 2e-05
• num_train_epochs: 1
• warmup_ratio: 0.1
• seed: 24
• 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: 128
• per_device_eval_batch_size: 128
• 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: 24
• 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: 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: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	gooaq-dev_cosine_ndcg@10
-1	-1	-	-	0.0000
0.0003	1	4.9236	-	-
0.0128	50	4.8759	-	-
0.0256	100	4.7225	-	-
0.0384	150	4.0357	-	-
0.0512	200	3.0877	-	-
0.0640	250	2.5094	-	-
0.0768	300	2.0771	-	-
0.0896	350	1.734	-	-
0.1024	400	1.4959	-	-
0.1152	450	1.308	-	-
0.1280	500	1.1529	0.8984	0.0796
0.1408	550	1.0101	-	-
0.1536	600	0.9601	-	-
0.1664	650	0.8845	-	-
0.1792	700	0.8348	-	-
0.1920	750	0.7838	-	-
0.2048	800	0.7457	-	-
0.2176	850	0.6879	-	-
0.2304	900	0.6778	-	-
0.2432	950	0.6783	-	-
0.2560	1000	0.6351	0.4814	0.2080
0.2687	1050	0.6221	-	-
0.2815	1100	0.6015	-	-
0.2943	1150	0.5738	-	-
0.3071	1200	0.5745	-	-
0.3199	1250	0.574	-	-
0.3327	1300	0.5464	-	-
0.3455	1350	0.5257	-	-
0.3583	1400	0.5074	-	-
0.3711	1450	0.4905	-	-
0.3839	1500	0.4633	0.3643	0.2435
0.3967	1550	0.4853	-	-
0.4095	1600	0.4587	-	-
0.4223	1650	0.4561	-	-
0.4351	1700	0.4442	-	-
0.4479	1750	0.4399	-	-
0.4607	1800	0.4448	-	-
0.4735	1850	0.4159	-	-
0.4863	1900	0.424	-	-
0.4991	1950	0.419	-	-
0.5119	2000	0.4049	0.3047	0.2713
0.5247	2050	0.3897	-	-
0.5375	2100	0.3873	-	-
0.5503	2150	0.3892	-	-
0.5631	2200	0.3777	-	-
0.5759	2250	0.382	-	-
0.5887	2300	0.3703	-	-
0.6015	2350	0.3703	-	-
0.6143	2400	0.3809	-	-
0.6271	2450	0.3576	-	-
0.6399	2500	0.3486	0.2686	0.2837
0.6527	2550	0.3395	-	-
0.6655	2600	0.3687	-	-
0.6783	2650	0.365	-	-
0.6911	2700	0.3553	-	-
0.7039	2750	0.3446	-	-
0.7167	2800	0.3396	-	-
0.7295	2850	0.3505	-	-
0.7423	2900	0.359	-	-
0.7551	2950	0.3239	-	-
0.7679	3000	0.3408	0.2474	0.2440
0.7807	3050	0.3217	-	-
0.7934	3100	0.3367	-	-
0.8062	3150	0.3479	-	-
0.8190	3200	0.3278	-	-
0.8318	3250	0.3203	-	-
0.8446	3300	0.2966	-	-
0.8574	3350	0.3298	-	-
0.8702	3400	0.3291	-	-
0.8830	3450	0.3199	-	-
0.8958	3500	0.3302	0.2363	0.2783
0.9086	3550	0.3124	-	-
0.9214	3600	0.3136	-	-
0.9342	3650	0.3327	-	-
0.9470	3700	0.3214	-	-
0.9598	3750	0.3214	-	-
0.9726	3800	0.3123	-	-
0.9854	3850	0.3185	-	-
0.9982	3900	0.2999	-	-
-1	-1	-	-	0.2882

Environmental Impact

Carbon emissions were measured using CodeCarbon.

• Energy Consumed: 0.057 kWh
• Carbon Emitted: 0.022 kg of CO2
• Hours Used: 0.212 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: 3.5.0.dev0
• Transformers: 4.49.0.dev0
• PyTorch: 2.5.0+cu121
• Accelerate: 1.3.0
• Datasets: 2.20.0
• Tokenizers: 0.21.0

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",
}

MultipleNegativesRankingLoss

@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}
}