tomaarsen/reranker-distilroberta-base-quora-duplicates

CrossEncoder based on distilbert/distilroberta-base

This is a Cross Encoder model finetuned from distilbert/distilroberta-base on the quora-duplicates dataset using the sentence-transformers library. It computes scores for pairs of texts, which can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

• Model Type: Cross Encoder
• Base model: distilbert/distilroberta-base
• Maximum Sequence Length: 514 tokens
• Training Dataset:
  • quora-duplicates
• Language: en

Model Sources

• Documentation: Sentence Transformers Documentation
• Documentation: Cross Encoder Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Cross Encoders on Hugging Face

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("sentence_transformers_model_id")
# Get scores for pairs...
pairs = [
    ['What is the step by step guide to invest in share market in india?', 'What is the step by step guide to invest in share market?'],
    ['What is the story of Kohinoor (Koh-i-Noor) Diamond?', 'What would happen if the Indian government stole the Kohinoor (Koh-i-Noor) diamond back?'],
    ['How can I increase the speed of my internet connection while using a VPN?', 'How can Internet speed be increased by hacking through DNS?'],
    ['Why am I mentally very lonely? How can I solve it?', 'Find the remainder when [math]23^{24}[/math] is divided by 24,23?'],
    ['Which one dissolve in water quikly sugar, salt, methane and carbon di oxide?', 'Which fish would survive in salt water?'],
]
scores = model.predict(pairs)
print(scores.shape)
# [5]

# ... or rank different texts based on similarity to a single text
ranks = model.rank(
    'What is the step by step guide to invest in share market in india?',
    [
        'What is the step by step guide to invest in share market?',
        'What would happen if the Indian government stole the Kohinoor (Koh-i-Noor) diamond back?',
        'How can Internet speed be increased by hacking through DNS?',
        'Find the remainder when [math]23^{24}[/math] is divided by 24,23?',
        'Which fish would survive in salt water?',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Classification

• Datasets: quora-duplicates-dev and quora-duplicates-test
• Evaluated with CEClassificationEvaluator

Metric	quora-duplicates-dev	quora-duplicates-test
accuracy	0.8938	0.8938
accuracy_threshold	0.5089	0.5091
f1	0.8612	0.8612
f1_threshold	0.3856	0.3858
precision	0.8183	0.8183
recall	0.9089	0.9089
average_precision	0.9203	0.9203

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: 1 characters mean: 59.15 characters max: 354 characters	min: 6 characters mean: 60.74 characters max: 399 characters	0: ~64.20% 1: ~35.80%

• Samples:

  sentence1	sentence2	label
  What are the features of the Indian caste system?	What triggers you the most when you play video games?	0
  What is the best place to learn Mandarin Chinese in Singapore?	What is the best place in Singapore for durian in December?	0
  What will be Hillary Clinton's India policy if she wins the election?	How would the bilateral relationship between India and the USA be under Hillary Clinton's presidency?	1

• Loss: BinaryCrossEntropyLoss

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: 11 characters mean: 57.9 characters max: 244 characters	min: 12 characters mean: 59.33 characters max: 221 characters	0: ~62.00% 1: ~38.00%

• Samples:

  sentence1	sentence2	label
  What is the step by step guide to invest in share market in india?	What is the step by step guide to invest in share market?	0
  What is the story of Kohinoor (Koh-i-Noor) Diamond?	What would happen if the Indian government stole the Kohinoor (Koh-i-Noor) diamond back?	0
  How can I increase the speed of my internet connection while using a VPN?	How can Internet speed be increased by hacking through DNS?	0

• Loss: BinaryCrossEntropyLoss

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
• bf16: 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: 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: True
• 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: None
• hub_always_push: False
• 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
• 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
• use_liger_kernel: False
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	quora-duplicates-dev_average_precision	quora-duplicates-test_average_precision
-1	-1	-	-	0.3711	-
0.0167	100	0.6574	-	-	-
0.0333	200	0.4804	-	-	-
0.0500	300	0.4406	-	-	-
0.0666	400	0.4208	-	-	-
0.0833	500	0.3929	0.3958	0.8210	-
0.0999	600	0.3986	-	-	-
0.1166	700	0.3743	-	-	-
0.1332	800	0.3938	-	-	-
0.1499	900	0.3602	-	-	-
0.1665	1000	0.3714	0.3437	0.8565	-
0.1832	1100	0.3486	-	-	-
0.1998	1200	0.3479	-	-	-
0.2165	1300	0.3417	-	-	-
0.2331	1400	0.3425	-	-	-
0.2498	1500	0.3353	0.3264	0.8742	-
0.2664	1600	0.3335	-	-	-
0.2831	1700	0.3274	-	-	-
0.2998	1800	0.3284	-	-	-
0.3164	1900	0.3118	-	-	-
0.3331	2000	0.3073	0.3282	0.8826	-
0.3497	2100	0.3233	-	-	-
0.3664	2200	0.3072	-	-	-
0.3830	2300	0.314	-	-	-
0.3997	2400	0.3065	-	-	-
0.4163	2500	0.3046	0.2877	0.8930	-
0.4330	2600	0.2857	-	-	-
0.4496	2700	0.285	-	-	-
0.4663	2800	0.2957	-	-	-
0.4829	2900	0.2965	-	-	-
0.4996	3000	0.2824	0.2842	0.8998	-
0.5162	3100	0.3019	-	-	-
0.5329	3200	0.2841	-	-	-
0.5495	3300	0.2981	-	-	-
0.5662	3400	0.2878	-	-	-
0.5828	3500	0.278	0.2803	0.9061	-
0.5995	3600	0.2841	-	-	-
0.6162	3700	0.2794	-	-	-
0.6328	3800	0.2808	-	-	-
0.6495	3900	0.27	-	-	-
0.6661	4000	0.2719	0.2697	0.9091	-
0.6828	4100	0.2792	-	-	-
0.6994	4200	0.2669	-	-	-
0.7161	4300	0.2696	-	-	-
0.7327	4400	0.2642	-	-	-
0.7494	4500	0.2684	0.2591	0.9140	-
0.7660	4600	0.2593	-	-	-
0.7827	4700	0.2756	-	-	-
0.7993	4800	0.2584	-	-	-
0.8160	4900	0.2525	-	-	-
0.8326	5000	0.267	0.2540	0.9168	-
0.8493	5100	0.2612	-	-	-
0.8659	5200	0.2607	-	-	-
0.8826	5300	0.2565	-	-	-
0.8993	5400	0.2432	-	-	-
0.9159	5500	0.2568	0.2489	0.9198	-
0.9326	5600	0.2572	-	-	-
0.9492	5700	0.2658	-	-	-
0.9659	5800	0.2568	-	-	-
0.9825	5900	0.2539	-	-	-
0.9992	6000	0.2458	0.2503	0.9203	-
-1	-1	-	-	-	0.9203

Environmental Impact

Carbon emissions were measured using CodeCarbon.

• Energy Consumed: 0.069 kWh
• Carbon Emitted: 0.027 kg of CO2
• Hours Used: 0.214 hours

Training Hardware

• On Cloud: No
• GPU Model: 1 x NVIDIA GeForce RTX 3090
• CPU Model: 13th Gen Intel(R) Core(TM) i7-13700K
• RAM Size: 31.78 GB

Framework Versions

• Python: 3.11.6
• Sentence Transformers: 3.5.0.dev0
• Transformers: 4.49.0.dev0
• PyTorch: 2.5.0+cu121
• Accelerate: 1.3.0
• Datasets: 2.20.0
• Tokenizers: 0.21.0

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",
}