metadata

language: []
library_name: sentence-transformers
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - dataset_size:100K<n<1M
  - loss:MultipleNegativesRankingLoss
base_model: FacebookAI/xlm-roberta-base
widget:
  - source_sentence: who does carol end up with on er
    sentences:
      - who does sarah end up with in must love dogs
      - when does season 7 once upon a time start
      - when did the wwe hall of fame start
  - source_sentence: who is mr blue in the hulk movie
    sentences:
      - who played the zither in the third man
      - who plays tiana s mom in once upon a time
      - when did 9 11 start being called patriot day
  - source_sentence: who do you play as in bioshock 1
    sentences:
      - who does brad pitt play in war machine
      - where does gone in 60 seconds take place
      - in 1864 the republicans changed their name to
  - source_sentence: when did the xbox one s came out
    sentences:
      - when did the first generation iphone come out
      - what is the meaning of cdi in motorcycle
      - who sang in the court of the crimson king
  - source_sentence: how big is 1 acre in square feet
    sentences:
      - how many square feet is one acre of land
      - what type of bonding is present in alloys
      - who wrote skinny love bon iver and birdy
pipeline_tag: sentence-similarity

SentenceTransformer based on FacebookAI/xlm-roberta-base

This is a sentence-transformers model finetuned from FacebookAI/xlm-roberta-base. 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: FacebookAI/xlm-roberta-base
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: XLMRobertaModel 
  (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("Stern5497/nir-24-xlm-roberta-base")
# Run inference
sentences = [
    'how big is 1 acre in square feet',
    'how many square feet is one acre of land',
    'what type of bonding is present in alloys',
]
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]

Training Details

Training Dataset

Unnamed Dataset

Size: 164,848 training samples
Columns: sentence_0, sentence_1, and sentence_2

Approximate statistics based on the first 1000 samples:

	sentence_0	sentence_1	sentence_2
type	string	string	string
details	min: 10 tokens mean: 13.28 tokens max: 29 tokens	min: 131 tokens mean: 163.32 tokens max: 231 tokens	min: 133 tokens mean: 164.04 tokens max: 290 tokens

Samples:

sentence_0	sentence_1	sentence_2
`when did the song say my name come out`	{'title': '', 'text': 'Say My Name "Say My Name" is a song by the American girl group Destiny's Child from their second studio album "The Writing's on the Wall" (1999). It was written by Beyoncé Knowles, LeToya Luckett, LaTavia Roberson and Kelly Rowland along with LaShawn Daniels, Fred Jerkins III and Rodney "Darkchild" Jerkins, featuring production by the latter. The song was released by Columbia Records as the album's third single on November 7, 1999, and marked the introduction of the group's second line-up with the newly-added members Michelle Williams and Farrah Franklin. The single is the most successful of the four releases'}	{'title': '', 'text': 'Romans 1 Romans 1 is the first chapter of the Epistle to the Romans in the New Testament of the Christian Bible. It is authored by Paul the Apostle, but written by an amanuensis, Tertius, while Paul was in Corinth, in winter of AD 57-58. records that Paul stayed in Greece, probably Corinth, for three months. Paul wrote to the Roman Christians in order to give them a substantial resume of his theology. The New King James Version organises this chapter as follows: The letter is addressed "to all in Rome who are loved by God and called to be'}
`what is another name for the persian empire`	{'title': '', 'text': 'in the neighbouring Ottoman Empire. Persian Empire The Persian Empire (, translit. "", lit. 'Imperial Iran') refers to any of a series of imperial dynasties that were centred in Persia/Iran from the 6th century BC Achaemenid Empire era to the 20th century AD in the Qajar dynasty era. The first dynasty of the Persian Empire was created by Achaemenids, established by Cyrus the Great in 550 BC with the conquest of Median, Lydian and Babylonian empires. It covered much of the Ancient world when it was conquered by Alexander the Great. Persepolis is the most famous historical site related to'}	{'title': '', 'text': 'squad leader is staff sergeant (E-6, or OR-6) and in the United States Marine Corps the TO rank is sergeant (E-5, or OR-5), though a corporal may also act as a squad leader in the absence of sufficient numbers of sergeants. Squad leaders of crew-served weapons squads range from corporal through staff sergeant, depending upon the branch of service and type of squad. In some armies, notably those of the British Commonwealth, in which the term section is used for units of this size, the NCO in charge, which in the British Army and Royal Marines is normally a Corporal'}
`what is the unit of measurement in australia`	{'title': '', 'text': 'Metrication in Australia Metrication in Australia effectively began in 1966 with the conversion to decimal currency under the auspices of the Decimal Currency Board. The conversion of measurements—metrication—commenced subsequently in 1971, under the direction of the Metric Conversion Board and actively proceeded until the Board was disbanded in 1981. Before 1970, Australia mostly used the imperial system for measurement, which the Australian colonies had inherited from the United Kingdom. Between 1970 and 1988, imperial units were withdrawn from general legal use and replaced with SI metric units, facilitated through legislation and government agencies. SI units are now the sole legal'}	{'title': '', 'text': 'squad of several teams, but did not see NFL action until the 1998 NFL season with the Miami Dolphins. He then played for the Green Bay Packers and Cleveland Browns, and was a member of the New York Giants but did not play due to back injuries. He spent time in NFL Europe, where he played for the Scottish Claymores. Barry Stokes (American football) Barry Wade Stokes (born December 20, 1973) is a former American football offensive lineman of the National Football League. He was signed by the Detroit Lions as an undrafted free agent in 1996. He played college'}

Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 1
fp16: True
batch_sampler: no_duplicates
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
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
num_train_epochs: 1
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
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}
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: round_robin

Training Logs

Epoch	Step	Training Loss
0.0485	500	1.5522
0.0971	1000	0.816
0.1456	1500	0.6951
0.1941	2000	0.6024
0.2426	2500	0.5464
0.2912	3000	0.5076
0.3397	3500	0.459
0.3882	4000	0.4633
0.4368	4500	0.4309
0.4853	5000	0.4081
0.5338	5500	0.3783
0.5824	6000	0.3624
0.6309	6500	0.3467
0.6794	7000	0.3358
0.7279	7500	0.3279
0.7765	8000	0.3062
0.8250	8500	0.3029
0.8735	9000	0.305
0.9221	9500	0.2927
0.9706	10000	0.272
1.0	10303	-

Framework Versions

Python: 3.10.13
Sentence Transformers: 3.0.0
Transformers: 4.39.3
PyTorch: 2.1.2
Accelerate: 0.29.3
Datasets: 2.18.0
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",
}

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}
}