omega5505/stsb-distilbert-base-ocl

SentenceTransformer based on sentence-transformers/stsb-distilbert-base

This is a sentence-transformers model finetuned from sentence-transformers/stsb-distilbert-base on the quora-duplicates 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/stsb-distilbert-base
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • quora-duplicates
• Language: en

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': 128, '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("omega5505/stsb-distilbert-base-ocl")
# Run inference
sentences = [
    'Why do so many religious people believe in healing miracles?',
    'Is believing in God a bad thing?',
    'What do you like about China?',
]
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

Binary Classification

• Dataset: quora-duplicates
• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.877
cosine_accuracy_threshold	0.7857
cosine_f1	0.8516
cosine_f1_threshold	0.7746
cosine_precision	0.8209
cosine_recall	0.8847
cosine_ap	0.8988
cosine_mcc	0.7484

Paraphrase Mining

• Dataset: quora-duplicates-dev
• Evaluated with ParaphraseMiningEvaluator

Metric	Value
average_precision	0.5483
f1	0.5606
precision	0.5539
recall	0.5675
threshold	0.8632

Information Retrieval

• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9308
cosine_accuracy@3	0.969
cosine_accuracy@5	0.9778
cosine_accuracy@10	0.9854
cosine_precision@1	0.9308
cosine_precision@3	0.4145
cosine_precision@5	0.267
cosine_precision@10	0.1414
cosine_recall@1	0.8009
cosine_recall@3	0.9314
cosine_recall@5	0.9558
cosine_recall@10	0.9744
cosine_ndcg@10	0.9511
cosine_mrr@10	0.9512
cosine_map@100	0.9391

Training Details

Training Dataset

quora-duplicates

• Dataset: quora-duplicates at 451a485
• Size: 404,290 training samples
• Columns: sentence1, sentence2, and label
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	int
  details	min: 6 tokens mean: 15.73 tokens max: 65 tokens	min: 6 tokens mean: 15.93 tokens max: 85 tokens	0: ~61.60% 1: ~38.40%

• Samples:

  sentence1	sentence2	label
  How can Trump supporters claim he didn't mock a disabled reporter when there is live footage of him mocking a disabled reporter?	Why don't people actually watch the Trump video of him allegedly mocking a disabled reporter?	0
  Where can I get the best digital marketing course (online & offline) in India?	Which is the best digital marketing institute for professionals in India?	1
  What best two liner shayri?	What does "senile dementia, uncomplicated" mean in medical terms?	0

• Loss: OnlineContrastiveLoss

Evaluation Dataset

quora-duplicates

• Dataset: quora-duplicates at 451a485
• Size: 404,290 evaluation samples
• Columns: sentence1, sentence2, and label
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	int
  details	min: 6 tokens mean: 16.14 tokens max: 70 tokens	min: 6 tokens mean: 15.92 tokens max: 74 tokens	0: ~60.10% 1: ~39.90%

• Samples:

  sentence1	sentence2	label
  What are some must subscribe RSS feeds?	What are RSS feeds?	0
  How close are Madonna and Hillary Clinton?	Why do people say Hillary Clinton is a crook?	0
  Can you share best day of your life?	What is the Best Day of your life till date?	1

• Loss: OnlineContrastiveLoss

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• num_train_epochs: 1
• warmup_ratio: 0.1
• fp16: 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: 64
• per_device_eval_batch_size: 64
• 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: 1.0
• 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: 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
• eval_on_start: False
• eval_use_gather_object: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	quora-duplicates_cosine_ap	quora-duplicates-dev_average_precision	cosine_ndcg@10
0	0	-	-	0.7458	0.4200	0.9390
0.0640	100	2.5263	-	-	-	-
0.1280	200	2.1489	-	-	-	-
0.1599	250	-	1.8621	0.8433	0.3907	0.9329
0.1919	300	2.0353	-	-	-	-
0.2559	400	1.7831	-	-	-	-
0.3199	500	1.8887	1.7744	0.8662	0.4924	0.9379
0.3839	600	1.7814	-	-	-	-
0.4479	700	1.7775	-	-	-	-
0.4798	750	-	1.6468	0.8766	0.4945	0.9399
0.5118	800	1.6835	-	-	-	-
0.5758	900	1.6974	-	-	-	-
0.6398	1000	1.5704	1.4925	0.8895	0.5283	0.9460
0.7038	1100	1.6771	-	-	-	-
0.7678	1200	1.619	-	-	-	-
0.7997	1250	-	1.4311	0.8982	0.5252	0.9466
0.8317	1300	1.6119	-	-	-	-
0.8957	1400	1.6043	-	-	-	-
0.9597	1500	1.6848	1.4070	0.8988	0.5483	0.9511

Framework Versions

• Python: 3.9.18
• Sentence Transformers: 3.4.1
• Transformers: 4.44.2
• PyTorch: 2.2.1+cu121
• Accelerate: 1.3.0
• Datasets: 2.19.0
• 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",
}