Rich740804/st-scale70

SentenceTransformer based on sentence-transformers/paraphrase-mpnet-base-v2

This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-mpnet-base-v2 on the train 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: sentence-transformers/paraphrase-mpnet-base-v2
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • train

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("sentence_transformers_model_id")
# Run inference
sentences = [
    'The marriage of Baptiste and Hannah demonstrates their commitment to sharing their lives and supporting one another.',
    'By getting married, Baptiste and Hannah take on a duty to care for each other, both emotionally and materially.',
    'If the marriage brings happiness to Baptiste and Hannah, then they are pursuing their right to happiness.',
]
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]

Training Details

Training Dataset

train

• Dataset: train
• Size: 18,963 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 10 tokens mean: 25.92 tokens max: 51 tokens	min: 9 tokens mean: 28.31 tokens max: 60 tokens	min: 11 tokens mean: 28.69 tokens max: 67 tokens

• Samples:

  anchor	positive	negative
  Saving the group of people from harm by diverting the trolley supports the value of preserving life.	The group of people tied to the tracks have a right to life, which is protected when the trolley is diverted to save them.	Diverting the trolley reduces overall harm by preventing the deaths of many people at the cost of one person's life.
  The bake sale could be seen as an expression of support for a particular cause, and the right to freely express oneself and associate with others who share the same views is important.	The bake sale might be seen as a form of protest or support for a specific cause, and individuals have the right to engage in peaceful protest or show support.	If the bake sale directly or indirectly promotes religious discrimination, this can infringe on the fundamental right of individuals to be free from discrimination or harm due to their religious beliefs.
  Children have a right to life, and saving them from danger upholds this right.	Children should be protected from harm, abuse, and danger, and saving them ensures this right is respected.	Children have a right to grow up with access to healthcare, education, and a nurturing environment. Saving them may help secure these rights.

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 40,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• overwrite_output_dir: True
• per_device_train_batch_size: 32
• learning_rate: 2.1456771788455288e-05
• num_train_epochs: 2
• warmup_ratio: 0.03254893834779507
• fp16: True
• dataloader_num_workers: 4
• remove_unused_columns: False

All Hyperparameters

Click to expand

• overwrite_output_dir: True
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 8
• 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: 2.1456771788455288e-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: 2
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.03254893834779507
• 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: True
• 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: 4
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: False
• 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
• 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
0.0337	20	0.2448
0.0675	40	0.1918
0.1012	60	0.14
0.1349	80	0.186
0.1686	100	0.1407
0.2024	120	0.1672
0.2361	140	0.1832
0.2698	160	0.116
0.3035	180	0.1341
0.3373	200	0.2118
0.3710	220	0.1274
0.4047	240	0.1993
0.4384	260	0.1561
0.4722	280	0.1517
0.5059	300	0.1635
0.5396	320	0.1646
0.5734	340	0.1337
0.6071	360	0.1406
0.6408	380	0.1114
0.6745	400	0.1314
0.7083	420	0.1481
0.7420	440	0.1932
0.7757	460	0.1568
0.8094	480	0.1319
0.8432	500	0.1536
0.8769	520	0.1462
0.9106	540	0.1336
0.9444	560	0.1453
0.9781	580	0.2005
1.0118	600	0.1265
1.0455	620	0.0702
1.0793	640	0.0739
1.1130	660	0.049
1.1467	680	0.0613
1.1804	700	0.0663
1.2142	720	0.0726
1.2479	740	0.0822
1.2816	760	0.0651
1.3153	780	0.0603
1.3491	800	0.0468
1.3828	820	0.061
1.4165	840	0.0891
1.4503	860	0.0607
1.4840	880	0.0673
1.5177	900	0.0728
1.5514	920	0.065
1.5852	940	0.0824
1.6189	960	0.0695
1.6526	980	0.0626
1.6863	1000	0.0525
1.7201	1020	0.0482
1.7538	1040	0.0968
1.7875	1060	0.0717
1.8212	1080	0.0704
1.8550	1100	0.0666
1.8887	1120	0.0841
1.9224	1140	0.0682
1.9562	1160	0.0584
1.9899	1180	0.0423

Framework Versions

• Python: 3.9.21
• Sentence Transformers: 4.1.0
• Transformers: 4.52.4
• PyTorch: 2.6.0+cu124
• Accelerate: 1.5.2
• Datasets: 3.4.1
• Tokenizers: 0.21.1

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}
}