This is a sentence-transformers model finetuned from sentence-transformers/LaBSE. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for retrieval.

Model was fine-tuned Ruska Romani on Russian language pairs. Ruska Romani is the dialect of Romani language attributed to Ruska Roma, the largest subgroup of Romani people in Russia.

Model was trained and evaluated on parallel pairs data data from the Russian National Corpus.

The data curation process is described in the article The Parallel Corpus of Russian and Ruska Romani Languages. Please refer to that paper in any publications where the model was used.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: sentence-transformers/LaBSE
Maximum Sequence Length: 256 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Supported Modality: Text

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'BertModel'})
  (1): Pooling({'embedding_dimension': 768, 'pooling_mode': 'cls', 'include_prompt': True})
  (2): Dense({'in_features': 768, 'out_features': 768, 'bias': True, 'activation_function': 'torch.nn.modules.activation.Tanh', 'module_input_name': 'sentence_embedding', 'module_output_name': 'sentence_embedding'})
  (3): Normalize({})
)

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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("midwestcyr/labse-ruskaromani")
# Run inference
sentences = [
    'Владимир разорвал их, не читая.',
    'Владимиро розрискирдя лэн на гины.',
    'Мандэ кэ ёв сыс баро уважэниё и ёв ман дрэван камья.',
]
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, 0.8380, 0.0814],
#         [0.8380, 1.0000, 0.1824],
#         [0.0814, 0.1824, 1.0000]])

Evaluation

Metrics

Translation

Dataset: ruska-roma-validation
Evaluated with TranslationEvaluator

Metric	Value
src2trg_accuracy	0.9023
trg2src_accuracy	0.8925
mean_accuracy	0.8974

Training Details

Training Dataset

Unnamed Dataset

Size: 7,278 training samples
Columns: sentence_0, sentence_1, and label
Approximate statistics based on the first 100 samples:
sentence_0 sentence_1 label
type string string float
modality text text
details
min: 5 tokens
mean: 23.74 tokens
max: 96 tokens
min: 6 tokens
mean: 33.36 tokens
max: 135 tokens
min: 1.0
mean: 1.0
max: 1.0

Samples:

sentence_0	sentence_1	label
`Не бойтесь, государь милостив, я буду просить его. Он нас не обидит. Мы все его дети. А как ему за вас будет заступиться, если вы станете бунтовать и разбойничать».`	`На дарэнте, тагари куч, мэ лава тэ мангав лэс — ёв амэн на помэкэла — амэ сарэ лэскирэ чявэ — а сыр лэскэ пал тумэндэ тэ затэрдёл, коли тумэ лэна тэ газдэн бунто и разбоё.`	`1.0`
`Можно было видеть, что мужчина высок и стоит у весла, широко расставив ноги, вполоборота к кругленькой, маленькой женщине, прислонившейся грудью к другому веслу, саженях в полутора от первого.`	`Могискирдо сыс тэ роздыкхэс, со гаджё учё и тэрдо пашо вёсло, буґлэс росчеви ґэра, дро паш обрисибэ кэ крэнглинько джювлы, сави припасия колынэса кэ вавир вёсло, надур екх екхэстыр.`	`1.0`
`Мазурка кончилась, хозяева просили гостей к ужину, но полковник Б. отказался, сказав, что ему надо завтра рано вставать, и простился с хозяевами.`	`Мазурка кончисалыя, хулая мангнэ гостен ко хабэ, нэ полковнико Б. отпхэндяпэ и пхэндя, со лэскэ трэбинэ атася злокоса тэ уштэс и простиндяпэ хуланца.`	`1.0`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim", 4 "gather_across_devices": false, 5 "directions": [ 6 "query_to_doc" 7 ], 8 "partition_mode": "joint", 9 "hardness_mode": null, 10 "hardness_strength": 0.0 11}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
fp16: True
per_device_eval_batch_size: 16
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

per_device_train_batch_size: 16
num_train_epochs: 3
max_steps: -1
learning_rate: 5e-05
lr_scheduler_type: linear
lr_scheduler_kwargs: None
warmup_steps: 0
optim: adamw_torch_fused
optim_args: None
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
optim_target_modules: None
gradient_accumulation_steps: 1
average_tokens_across_devices: True
max_grad_norm: 1
label_smoothing_factor: 0.0
bf16: False
fp16: True
bf16_full_eval: False
fp16_full_eval: False
tf32: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
use_liger_kernel: False
liger_kernel_config: None
use_cache: False
neftune_noise_alpha: None
torch_empty_cache_steps: None
auto_find_batch_size: False
log_on_each_node: True
logging_nan_inf_filter: True
include_num_input_tokens_seen: no
log_level: passive
log_level_replica: warning
disable_tqdm: False
project: huggingface
trackio_space_id: None
trackio_bucket_id: None
trackio_static_space_id: None
per_device_eval_batch_size: 16
prediction_loss_only: True
eval_on_start: False
eval_do_concat_batches: True
eval_use_gather_object: False
eval_accumulation_steps: None
include_for_metrics: []
batch_eval_metrics: False
save_only_model: False
save_on_each_node: False
enable_jit_checkpoint: False
push_to_hub: False
hub_private_repo: None
hub_model_id: None
hub_strategy: every_save
hub_always_push: False
hub_revision: None
load_best_model_at_end: False
ignore_data_skip: False
restore_callback_states_from_checkpoint: False
full_determinism: False
seed: 42
data_seed: None
use_cpu: False
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
parallelism_config: None
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_pin_memory: True
dataloader_persistent_workers: False
dataloader_prefetch_factor: None
remove_unused_columns: True
label_names: None
train_sampling_strategy: random
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
ddp_static_graph: None
ddp_backend: None
ddp_timeout: 1800
fsdp: None
fsdp_config: None
deepspeed: None
debug: []
skip_memory_metrics: True
do_predict: False
resume_from_checkpoint: None
warmup_ratio: None
local_rank: -1
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	ruska-roma-validation_mean_accuracy
1.0	455	-	0.8535
1.0989	500	0.1608	0.8684
2.0	910	-	0.8875
2.1978	1000	0.0325	0.8956
3.0	1365	-	0.8974

Training Time

Training: 11.7 minutes

Framework Versions

Python: 3.12.13
Sentence Transformers: 5.6.0
Transformers: 5.12.1
PyTorch: 2.11.0+cu128
Accelerate: 1.14.0
Datasets: 4.0.0
Tokenizers: 0.22.2

Citation

BibTeX

Ruska Romani and Russian Parallel Corpus

bibtex

@inproceedings{koncha-etal-2024-parallel,
    title = "The Parallel Corpus of {R}ussian and Ruska {R}omani Languages",
    author = "Koncha, Kirill  and
      Kukanova, Abina  and
      Tatiana, Kazakova  and
      Rozovskaya, Gloria",
    editor = "Serikov, Oleg  and
      Voloshina, Ekaterina  and
      Postnikova, Anna  and
      Muradoglu, Saliha  and
      Le Ferrand, Eric  and
      Klyachko, Elena  and
      Vylomova, Ekaterina  and
      Shavrina, Tatiana  and
      Tyers, Francis",
    booktitle = "Proceedings of the Third Workshop on NLP Applications to Field Linguistics",
    month = aug,
    year = "2024",
    address = "Bangkok, Thailand",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.fieldmatters-1.1/",
    doi = "10.18653/v1/2024.fieldmatters-1.1",
    pages = "1--5"
}

Sentence Transformers

bibtex

@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

bibtex

@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}

	sentence_0	sentence_1	label
type	string	string	float
modality	text	text
details	min: 5 tokens mean: 23.74 tokens max: 96 tokens	min: 6 tokens mean: 33.36 tokens max: 135 tokens	min: 1.0 mean: 1.0 max: 1.0