hcy5561/distilroberta-base-sentence-transformer-triplets

SentenceTransformer based on google-bert/bert-base-uncased

This is a sentence-transformers model finetuned from google-bert/bert-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: google-bert/bert-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: BertModel 
  (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("hcy5561/distilroberta-base-sentence-transformer-triplets")
# Run inference
sentences = [
    'How did Halloween Originate? What country did it originate on?',
    'In what country did Halloween originate?',
    'What was Halloween like in the 1990s?',
]
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

Triplet

• Dataset: QQP-nli-dev
• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9878
dot_accuracy	0.0124
manhattan_accuracy	0.9874
euclidean_accuracy	0.9878
max_accuracy	0.9878

Training Details

Training Dataset

Unnamed Dataset

• Size: 91,585 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 6 tokens mean: 13.95 tokens max: 50 tokens	min: 6 tokens mean: 14.02 tokens max: 52 tokens	min: 6 tokens mean: 14.68 tokens max: 60 tokens

• Samples:

  anchor	positive	negative
  How can I overcome a bad mood?	How do I break out of a bad mood?	The world around me seems so austere and gloomy because of my mood. It's depressing me considerably. What can I do?
  What are symptoms of mild schizophrenia?	What are some symptoms of when you become schizophrenic?	Is confusion another symptom of being schizophrenic?
  What are some ideas which transformed ordinary people into millionaires?	What are some things ordinary people know but millionaires don't?	What can billionaires do that millionaire cannot do?

• Loss: TripletLoss with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
      "triplet_margin": 5
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 5,088 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 6 tokens mean: 14.14 tokens max: 44 tokens	min: 6 tokens mean: 13.96 tokens max: 49 tokens	min: 6 tokens mean: 14.8 tokens max: 60 tokens

• Samples:

  anchor	positive	negative
  Why do I see the exact same questions in my feed all the time?	Why are too many questions repeating in my feed sometimes?	Why does this "question" keep showing up in the Unorganized Questions global_feed? (see description for screenshot)
  Can we expect time travel to become a reality?	Can we time travel anyhow?	What do you hAve to say about time travel (I am not science student but I read it on net and its so exciting topic but still no clear idea that is it possible or it's just a rumour)?
  Is it too late to start medical school at 32?	Is it too late to go to medical school at 24?	As a 14 year old girl who wants to go to medical school, should I work extremely hard and study a lot now to be ready for it? What should I do?

• Loss: TripletLoss with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
      "triplet_margin": 5
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• num_train_epochs: 4
• warmup_ratio: 0.1
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• 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
• 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
• 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}
• 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
• 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_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	QQP-nli-dev_max_accuracy
0	0	-	-	0.8783
0.1746	500	2.3079	0.8664	0.9581
0.3493	1000	0.9367	0.5027	0.9737
0.5239	1500	0.6747	0.4471	0.9743
0.6986	2000	0.5323	0.3740	0.9776
0.8732	2500	0.4765	0.3178	0.9825
1.0479	3000	0.4104	0.2809	0.9866
1.2225	3500	0.3266	0.2633	0.9870
1.3971	4000	0.2129	0.2566	0.9862
1.5718	4500	0.1559	0.2542	0.9858
1.7464	5000	0.1432	0.2482	0.9853
1.9211	5500	0.1361	0.2370	0.9845
2.0957	6000	0.1179	0.2102	0.9880
2.2703	6500	0.0921	0.2201	0.9870
2.4450	7000	0.0656	0.2075	0.9878
2.6196	7500	0.0497	0.2011	0.9876
2.7943	8000	0.0455	0.1960	0.9878
2.9689	8500	0.0422	0.1973	0.9872
3.1436	9000	0.0349	0.1863	0.9890
3.3182	9500	0.0319	0.1850	0.9882
3.4928	10000	0.02	0.1854	0.9882
3.6675	10500	0.0184	0.1849	0.9884
3.8421	11000	0.0178	0.1828	0.9878

Framework Versions

• Python: 3.10.6
• Sentence Transformers: 3.0.1
• Transformers: 4.39.3
• PyTorch: 2.2.2+cu118
• Accelerate: 0.28.0
• Datasets: 2.20.0
• Tokenizers: 0.15.2

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",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification}, 
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}