pierreinalco/custom-v1

SentenceTransformer based on distilbert/distilbert-base-uncased

This is a sentence-transformers model finetuned from distilbert/distilbert-base-uncased. 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: distilbert/distilbert-base-uncased
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• 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': 512, 'do_lower_case': False}) with Transformer model: DistilBertModel 
  (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 = [
    'Salinity gradients in oceans affect local wildlife habitats.',
    'The distribution of wildlife in different habitats has fascinated ecologists for decades.',
    '[SYNTAX] Bioenergy plants can convert agricultural waste into valuable electricity.',
]
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

• Dataset: custom-dev
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.9117
spearman_cosine	0.8442
pearson_manhattan	0.9157
spearman_manhattan	0.8441
pearson_euclidean	0.916
spearman_euclidean	0.8446
pearson_dot	0.9046
spearman_dot	0.8328
pearson_max	0.916
spearman_max	0.8446

Semantic Similarity

• Dataset: custom-test
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.9198
spearman_cosine	0.8501
pearson_manhattan	0.9282
spearman_manhattan	0.8494
pearson_euclidean	0.9284
spearman_euclidean	0.8498
pearson_dot	0.9141
spearman_dot	0.8411
pearson_max	0.9284
spearman_max	0.8501

Training Details

Training Dataset

Unnamed Dataset

• Size: 19,352 training samples
• Columns: s1, s2, and label
• Approximate statistics based on the first 1000 samples:

  	s1	s2	label
  type	string	string	int
  details	min: 10 tokens mean: 19.92 tokens max: 42 tokens	min: 10 tokens mean: 20.53 tokens max: 42 tokens	0: ~50.50% 1: ~49.50%

• Samples:

  s1	s2	label
  According to labeling theory, individuals are considered deviant once society has tagged them with that label.	Labeling theory posits that corporations become powerful when labeled as such by stakeholders.	0
  Employers must classify workers correctly as either employees or independent contractors to comply with tax and labor laws.	Employers must classify workers correctly as either employees or independent contractors to comply with tax and labor laws.	1
  Higher education institutions play a critical role in advancing research and innovation.	Advancement in research and innovation is significantly driven by the contributions of higher education institutions.	1

• Loss: CosineSimilarityLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 2,419 evaluation samples
• Columns: s1, s2, and label
• Approximate statistics based on the first 1000 samples:

  	s1	s2	label
  type	string	string	int
  details	min: 11 tokens mean: 19.91 tokens max: 37 tokens	min: 11 tokens mean: 20.46 tokens max: 42 tokens	0: ~49.70% 1: ~50.30%

• Samples:

  s1	s2	label
  Acoustic tomography is an innovative geophysical technique used to image the Earth's interior.	Acoustic tomography is an innovative geophysical technique used to image the Earth's interior.	1
  Urban areas frequently exhibit a different age distribution pattern compared to rural areas.	Urban areas frequently exhibit a different age distribution pattern compared to rural areas.	1
  Radiocarbon dating is a critical tool for assessing the duration of battery life in modern electronic devices.	Radiocarbon dating is a critical tool for assessing the duration of battery life in modern electronic devices.	1

• 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: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 10
• warmup_ratio: 0.1
• fp16: 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
• learning_rate: 5e-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: 10
• 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: 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: 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
• 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
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	custom-dev_spearman_cosine	custom-test_spearman_cosine
0.3300	100	0.2961	0.1185	0.8063	-
0.6601	200	0.0772	0.0504	0.8461	-
0.9901	300	0.0502	0.0454	0.8486	-
1.3201	400	0.0376	0.0402	0.8481	-
1.6502	500	0.0344	0.0400	0.8501	-
1.9802	600	0.0329	0.0390	0.8518	-
2.3102	700	0.0185	0.0387	0.8496	-
2.6403	800	0.0164	0.0371	0.8492	-
2.9703	900	0.0179	0.0393	0.8428	-
3.3003	1000	0.0099	0.0389	0.8466	-
3.6304	1100	0.0092	0.0395	0.8480	-
3.9604	1200	0.0101	0.0368	0.8492	-
4.2904	1300	0.0067	0.0385	0.8474	-
4.6205	1400	0.0056	0.0393	0.8456	-
4.9505	1500	0.0068	0.0401	0.8466	-
5.2805	1600	0.0041	0.0410	0.8462	-
5.6106	1700	0.0043	0.0399	0.8469	-
5.9406	1800	0.0039	0.0406	0.8463	-
6.2706	1900	0.003	0.0400	0.8456	-
6.6007	2000	0.0026	0.0416	0.8438	-
6.9307	2100	0.0027	0.0420	0.8437	-
7.2607	2200	0.0028	0.0424	0.8449	-
7.5908	2300	0.0021	0.0422	0.8458	-
7.9208	2400	0.002	0.0414	0.8451	-
8.2508	2500	0.0015	0.0421	0.8451	-
8.5809	2600	0.0015	0.0427	0.8451	-
8.9109	2700	0.0016	0.0429	0.8444	-
9.2409	2800	0.0011	0.0432	0.8442	-
9.5710	2900	0.0014	0.0432	0.8444	-
9.9010	3000	0.0011	0.0432	0.8442	-
10.0	3030	-	-	-	0.8501

Framework Versions

• Python: 3.11.9
• Sentence Transformers: 3.0.0
• Transformers: 4.41.2
• PyTorch: 2.3.0+cu121
• Accelerate: 0.30.1
• Datasets: 2.19.1
• Tokenizers: 0.19.1

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",
}