trbeers/distilbert-base-uncased-sts

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("trbeers/distilbert-base-uncased-sts")
# Run inference
sentences = [
    'Knowledge of medical equipment and veterinary terminology is necessary.',
    'Worked as a pet trainer for obedience classes',
    'Skilled in component sorting for various projects',
]
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: sts-dev
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.9243
spearman_cosine	0.8484
pearson_manhattan	0.9053
spearman_manhattan	0.8466
pearson_euclidean	0.9058
spearman_euclidean	0.8467
pearson_dot	0.9171
spearman_dot	0.8473
pearson_max	0.9243
spearman_max	0.8484

Semantic Similarity

• Dataset: sts-test
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.9188
spearman_cosine	0.8447
pearson_manhattan	0.8976
spearman_manhattan	0.8409
pearson_euclidean	0.8981
spearman_euclidean	0.8413
pearson_dot	0.9109
spearman_dot	0.8439
pearson_max	0.9188
spearman_max	0.8447

Training Details

Training Dataset

Unnamed Dataset

• Size: 8,137 training samples
• Columns: sentence1, sentence2, and score
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	score
  type	string	string	int
  details	min: 6 tokens mean: 16.34 tokens max: 40 tokens	min: 5 tokens mean: 9.58 tokens max: 24 tokens	0: ~49.50% 1: ~50.50%

• Samples:

  sentence1	sentence2	score
  Ability to use tools such as power drills as required for the job.	Proficient in operating power tools for installation tasks	1
  Experience with networking, specifically the TCP/IP stack, routing, ports, and services is essential.	Designed user interfaces for web applications	0
  Ability to establish and maintain positive relationships with coaches, student-athletes, and vendors regarding equipment selection.	Developed strong partnerships with vendors forEquipment procurement	1

• Loss: CosineSimilarityLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 2,035 evaluation samples
• Columns: sentence1, sentence2, and score
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	score
  type	string	string	int
  details	min: 6 tokens mean: 15.77 tokens max: 34 tokens	min: 5 tokens mean: 9.65 tokens max: 21 tokens	0: ~48.10% 1: ~51.90%

• Samples:

  sentence1	sentence2	score
  Experience with vulnerability management tools like Nessus and Nexpose.	managed network configurations	0
  Willingness to obtain a Texas fire extinguishers license as necessary.	Currently pursuing a Texas fire extinguishers license	1
  Experience in defining and maintaining enterprise architecture that supports business scalability.	Led the development of enterprise architecture frameworks for a multinational corporation	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: 4
• warmup_ratio: 0.1

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: 4
• 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: 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
• 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	sts-dev_spearman_cosine	sts-test_spearman_cosine
0.1965	100	0.1588	0.0884	0.8247	-
0.3929	200	0.0784	0.0686	0.8397	-
0.5894	300	0.067	0.0538	0.8455	-
0.7859	400	0.0626	0.0482	0.8450	-
0.9823	500	0.0533	0.0452	0.8454	-
1.1788	600	0.0346	0.0437	0.8434	-
1.3752	700	0.0328	0.0435	0.8465	-
1.5717	800	0.0306	0.0445	0.8465	-
1.7682	900	0.0317	0.0399	0.8481	-
1.9646	1000	0.0315	0.0448	0.8517	-
2.1611	1100	0.017	0.0388	0.8489	-
2.3576	1200	0.016	0.0396	0.8501	-
2.5540	1300	0.0129	0.0393	0.8465	-
2.7505	1400	0.0128	0.0396	0.8471	-
2.9470	1500	0.0147	0.0388	0.8483	-
3.1434	1600	0.009	0.0396	0.8460	-
3.3399	1700	0.0078	0.0390	0.8460	-
3.5363	1800	0.0063	0.0380	0.8475	-
3.7328	1900	0.0079	0.0377	0.8484	-
3.9293	2000	0.0062	0.0376	0.8484	-
4.0	2036	-	-	-	0.8447

Framework Versions

• Python: 3.10.11
• Sentence Transformers: 3.0.1
• Transformers: 4.41.2
• PyTorch: 2.3.1
• Accelerate: 0.31.0
• 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",
}