SentenceTransformer based on abkimc/distilroberta-base-sentence-transformer

This is a sentence-transformers model finetuned from abkimc/distilroberta-base-sentence-transformer. 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: abkimc/distilroberta-base-sentence-transformer
• Maximum Sequence Length: 512 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': 512, 'do_lower_case': False, 'architecture': 'RobertaModel'})
  (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("abkimc/distilroberta-base-sentence-transformer")
# Run inference
sentences = [
    'The HTC Legend has made its official debut in India days after it was informally launched .',
    'HTC Legend makes official debut in India',
    'Britain, Bill Gates join forces',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[ 1.0000,  0.9061, -0.0382],
#         [ 0.9061,  1.0000, -0.0170],
#         [-0.0382, -0.0170,  1.0000]])

Training Details

Training Dataset

Unnamed Dataset

• Size: 180,000 training samples
• Columns: sentence_0 and sentence_1
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1
  type	string	string
  details	min: 12 tokens mean: 33.68 tokens max: 293 tokens	min: 5 tokens mean: 10.98 tokens max: 28 tokens

• Samples:

  sentence_0	sentence_1
  Content is the king in today's world of journalism and a newspaper cannot survive if it compromises on the quality of the content, said Abhilash Khandekar, Maharashtra state head of Dainik Bhaskar Group on Tuesday.	'Content is king in today's journalism'
  Sammons Pensions has launched its ninth annual salary survey which aims to document remuneration packages across the industry.	Sammons launches ninth salary survey
  The state of Tennessee saw a major spike in foreclosure filings in 2008, according to a report by the Tennessee Housing Development Agency.	Tennessee sees major spike in foreclosure filings in 2008

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• num_train_epochs: 10
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• 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
• 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_fused
• 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
• hub_revision: None
• 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
• 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
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.1777	500	2.8662
0.3555	1000	0.0631
0.5332	1500	0.0149
0.7110	2000	0.0097
0.8887	2500	0.0079
1.0665	3000	0.0062
1.2442	3500	0.0041
1.4220	4000	0.0037
1.5997	4500	0.0038
1.7775	5000	0.0034
1.9552	5500	0.0038
2.1330	6000	0.0021
2.3107	6500	0.0015
2.4884	7000	0.0016
2.6662	7500	0.0015
2.8439	8000	0.0018
3.0217	8500	0.0015
3.1994	9000	0.0013
3.3772	9500	0.001
3.5549	10000	0.0011
3.7327	10500	0.0011
3.9104	11000	0.0014
4.0882	11500	0.0011
4.2659	12000	0.0007
4.4437	12500	0.0009
4.6214	13000	0.0009
4.7991	13500	0.0008
4.9769	14000	0.0008
5.1546	14500	0.0009
5.3324	15000	0.0007
5.5101	15500	0.0007
5.6879	16000	0.0007
5.8656	16500	0.0006
6.0434	17000	0.0007
6.2211	17500	0.0007
6.3989	18000	0.0005
6.5766	18500	0.0007
6.7544	19000	0.0005
6.9321	19500	0.0005
7.1098	20000	0.0005
7.2876	20500	0.0006
7.4653	21000	0.0005
7.6431	21500	0.0004
7.8208	22000	0.0004
7.9986	22500	0.0004
8.1763	23000	0.0004
8.3541	23500	0.0004
8.5318	24000	0.0005
8.7096	24500	0.0004
8.8873	25000	0.0004
9.0651	25500	0.0005
9.2428	26000	0.0004
9.4205	26500	0.0005
9.5983	27000	0.0004
9.7760	27500	0.0004
9.9538	28000	0.0004

Framework Versions

• Python: 3.12.11
• Sentence Transformers: 5.1.0
• Transformers: 4.55.4
• PyTorch: 2.8.0+cu126
• Accelerate: 1.10.1
• Datasets: 4.0.0
• Tokenizers: 0.21.4

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",
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}