README.md

metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:5749
  - loss:CosineSimilarityLoss
base_model: sentence-transformers/all-mpnet-base-v2
widget:
  - source_sentence: Young woman in riding gear on top of horse.
    sentences:
      - Italy’s centre-left splinters in presidential vote
      - The woman is riding on the brown horse.
      - Mali's Interim President Sworn Into Office
  - source_sentence: Sony reports record annual loss
    sentences:
      - A woman is playing a flute.
      - A man and a woman kiss.
      - Sony forecasts record annual loss of $6.4bn
  - source_sentence: A clear plastic chair in front of a bookcase.
    sentences:
      - Allen defends self against Farrow's abuse claims
      - Ehud Olmert sentenced to six years in Israel
      - a clear plastic chair in front of book shelves.
  - source_sentence: KLCI Futures traded mixed at mid-day
    sentences:
      - KL shares mixed at mid-day
      - NATO helicopter makes hard landing in E. Afghanistan
      - Sewol ferry crew faces trial
  - source_sentence: We in Britain think differently to Americans.
    sentences:
      - south korea has had a bullet train system since the 1980s.
      - Originally Posted by zaf We in Britain think differently to Americans.
      - Car bombings kill 13 civilians in Iraqi capital
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - pearson_cosine
  - spearman_cosine
model-index:
  - name: SentenceTransformer based on sentence-transformers/all-mpnet-base-v2
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: pearson_cosine
            value: 0.9075334661878893
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.9060484206473507
            name: Spearman Cosine
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: sts dev
          type: sts-dev
        metrics:
          - type: pearson_cosine
            value: 0.9075334589342524
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.9060484206473507
            name: Spearman Cosine

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

This is a sentence-transformers model finetuned from sentence-transformers/all-mpnet-base-v2. 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/all-mpnet-base-v2
• Maximum Sequence Length: 384 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity

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': 384, '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})
  (2): Normalize()
)

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 = [
    'We in Britain think differently to Americans.',
    'Originally Posted by zaf We in Britain think differently to Americans.',
    'south korea has had a bullet train system since the 1980s.',
]
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

Semantic Similarity

• Datasets: `` and sts-dev
• Evaluated with EmbeddingSimilarityEvaluator

Metric		sts-dev
pearson_cosine	0.9075	0.9075
spearman_cosine	0.906	0.906

Training Details

Training Dataset

Unnamed Dataset

• Size: 5,749 training samples
• Columns: sentence_0, sentence_1, and label
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	label
  type	string	string	float
  details	min: 6 tokens mean: 14.16 tokens max: 45 tokens	min: 5 tokens mean: 14.18 tokens max: 49 tokens	min: 0.0 mean: 0.54 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  US Senate to vote on fiscal cliff deal as deadline nears	Fiscal cliff: House delays vote on fiscal cliff deal - live	0.5599999904632569
  This is America, my friends, and it should not happen here," he said to loud applause.	"This is America, my friends, and it should not happen here."	0.65
  Books To Help Kids Talk About Boston Marathon News	Report of two explosions at finish line of Boston Marathon	0.1600000023841858

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• num_train_epochs: 10
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• 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: 1
• eval_accumulation_steps: None
• torch_empty_cache_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
• num_train_epochs: 10
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• 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: 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: batch_sampler
• multi_dataset_batch_sampler: round_robin

Training Logs

Click to expand

Epoch	Step	Training Loss	spearman_cosine	sts-dev_spearman_cosine
0	0	-	0.8811	-
0.1	18	-	-	0.8816
0.2	36	-	-	0.8834
0.3	54	-	-	0.8847
0.4	72	-	-	0.8894
0.5	90	-	-	0.8933
0.6	108	-	-	0.8966
0.7	126	-	-	0.9005
0.8	144	-	-	0.9020
0.9	162	-	-	0.9010
1.0	180	-	-	0.9001
1.1	198	-	-	0.9022
1.2	216	-	-	0.9018
1.3	234	-	-	0.9015
1.4	252	-	-	0.9029
1.5	270	-	-	0.9044
1.6	288	-	-	0.9049
1.7	306	-	-	0.9051
1.8	324	-	-	0.9033
1.9	342	-	-	0.9039
2.0	360	-	-	0.9050
2.1	378	-	-	0.9042
2.2	396	-	-	0.9041
2.3	414	-	-	0.9040
2.4	432	-	-	0.9048
2.5	450	-	-	0.9045
2.6	468	-	-	0.9046
2.7	486	-	-	0.9047
2.7778	500	0.0153	-	-
2.8	504	-	-	0.9057
2.9	522	-	-	0.9065
3.0	540	-	-	0.9074
3.1	558	-	-	0.9073
3.2	576	-	-	0.9065
3.3	594	-	-	0.9046
3.4	612	-	-	0.9057
3.5	630	-	-	0.9069
3.6	648	-	-	0.9062
3.7	666	-	-	0.9061
3.8	684	-	-	0.9050
3.9	702	-	-	0.9050
4.0	720	-	-	0.9048
4.1	738	-	-	0.9052
4.2	756	-	-	0.9055
4.3	774	-	-	0.9060
4.4	792	-	-	0.9059
4.5	810	-	-	0.9064
4.6	828	-	-	0.9063
4.7	846	-	-	0.9063
4.8	864	-	-	0.9067
4.9	882	-	-	0.9059
5.0	900	-	-	0.9052
5.1	918	-	-	0.9061
5.2	936	-	-	0.9057
5.3	954	-	-	0.9053
5.4	972	-	-	0.9060
5.5	990	-	-	0.9050
5.5556	1000	0.0051	-	-
5.6	1008	-	-	0.9053
5.7	1026	-	-	0.9052
5.8	1044	-	-	0.9056
5.9	1062	-	-	0.9062
6.0	1080	-	-	0.9056
6.1	1098	-	-	0.9054
6.2	1116	-	-	0.9058
6.3	1134	-	-	0.9058
6.4	1152	-	-	0.9056
6.5	1170	-	-	0.9057
6.6	1188	-	-	0.9055
6.7	1206	-	-	0.9055
6.8	1224	-	-	0.9053
6.9	1242	-	-	0.9053
7.0	1260	-	-	0.9053
7.1	1278	-	-	0.9057
7.2	1296	-	-	0.9055
7.3	1314	-	-	0.9053
7.4	1332	-	-	0.9056
7.5	1350	-	-	0.9059
7.6	1368	-	-	0.9060
7.7	1386	-	-	0.9057
7.8	1404	-	-	0.9058
7.9	1422	-	-	0.9057
8.0	1440	-	-	0.9058
8.1	1458	-	-	0.9059
8.2	1476	-	-	0.9060
8.3	1494	-	-	0.9056
8.3333	1500	0.0031	-	-
8.4	1512	-	-	0.9057
8.5	1530	-	-	0.9060
8.6	1548	-	-	0.9058
8.7	1566	-	-	0.9060
8.8	1584	-	-	0.9062
8.9	1602	-	-	0.9061
9.0	1620	-	-	0.9061
9.1	1638	-	-	0.9061
9.2	1656	-	-	0.9059
9.3	1674	-	-	0.9060
9.4	1692	-	-	0.9061
9.5	1710	-	-	0.9061
9.6	1728	-	-	0.9061
9.7	1746	-	-	0.9060
9.8	1764	-	-	0.9061
9.9	1782	-	-	0.9061
10.0	1800	-	0.9060	0.9060

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.3.1
• Transformers: 4.47.1
• PyTorch: 2.5.1+cu121
• Accelerate: 1.2.1
• Datasets: 3.2.0
• 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",
}