lingtrain / labse-chuvash

SentenceTransformer based on sentence-transformers/LaBSE

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 semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

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

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': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Dense({'in_features': 768, 'out_features': 768, 'bias': True, 'activation_function': 'torch.nn.modules.activation.Tanh'})
  (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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'Генри Джастис Форд',
    'Форд, Генри Джастис',
    'Я вышел из ванны свеж и бодр, как будто собирался на бал.',
]
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: 1,000,000 training samples
• Columns: sentence_0 , sentence_1 , and label
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	label
  type	string	string	float
  details	min: 3 tokens mean: 21.82 tokens max: 127 tokens	min: 4 tokens mean: 21.16 tokens max: 136 tokens	min: 1.0 mean: 1.0 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  Темех мар.	Дело десятое.	1.0
  Уругвайӑн тĕн ĕҫченĕсем	Религиозные деятели Уругвая	1.0
  Эп аванах ас тӑватӑп, пилӗк ҫул каялла пахчана эпир лайӑх тасатнӑччӗ.	А пять лет тому назад я знал, что сад был чищен.	1.0

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy : steps
• per_device_train_batch_size : 12
• per_device_eval_batch_size : 12
• num_train_epochs : 1
• fp16 : True
• multi_dataset_batch_sampler : round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir : False
• do_predict : False
• eval_strategy : steps
• prediction_loss_only : True
• per_device_train_batch_size : 12
• per_device_eval_batch_size : 12
• 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 : 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
• 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 : 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}
• tp_size : 0
• 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
• 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 : round_robin

Training Logs

Click to expand

Epoch	Step	Training Loss
0.0012	100	-
0.0024	200	-
0.0036	300	-
0.0048	400	-
0.0060	500	0.5331
0.0072	600	-
0.0084	700	-
0.0096	800	-
0.0108	900	-
0.0120	1000	0.3694
0.0132	1100	-
0.0144	1200	-
0.0156	1300	-
0.0168	1400	-
0.0180	1500	0.3141
0.0192	1600	-
0.0204	1700	-
0.0216	1800	-
0.0228	1900	-
0.0240	2000	0.2836
0.0252	2100	-
0.0264	2200	-
0.0276	2300	-
0.0288	2400	-
0.0300	2500	0.2823
0.0312	2600	-
0.0324	2700	-
0.0336	2800	-
0.0348	2900	-
0.0360	3000	0.265
0.0372	3100	-
0.0384	3200	-
0.0396	3300	-
0.0408	3400	-
0.0420	3500	0.2599
0.0432	3600	-
0.0444	3700	-
0.0456	3800	-
0.0468	3900	-
0.0480	4000	0.234
0.0492	4100	-
0.0504	4200	-
0.0516	4300	-
0.0528	4400	-
0.0540	4500	0.1966
0.0552	4600	-
0.0564	4700	-
0.0576	4800	-
0.0588	4900	-
0.0600	5000	0.2204
0.0612	5100	-
0.0624	5200	-
0.0636	5300	-
0.0648	5400	-
0.0660	5500	0.2272
0.0672	5600	-
0.0684	5700	-
0.0696	5800	-
0.0708	5900	-
0.0720	6000	0.2256
0.0732	6100	-
0.0744	6200	-
0.0756	6300	-
0.0768	6400	-
0.0780	6500	0.2071
0.0792	6600	-
0.0804	6700	-
0.0816	6800	-
0.0828	6900	-
0.0840	7000	0.2113
0.0852	7100	-
0.0864	7200	-
0.0876	7300	-
0.0888	7400	-
0.0900	7500	0.2222
0.0912	7600	-
0.0924	7700	-
0.0936	7800	-
0.0948	7900	-
0.0960	8000	0.2186
0.0972	8100	-
0.0984	8200	-
0.0996	8300	-
0.1008	8400	-
0.1020	8500	0.2137
0.1032	8600	-
0.1044	8700	-
0.1056	8800	-
0.1068	8900	-
0.1080	9000	0.1928
0.1092	9100	-
0.1104	9200	-
0.1116	9300	-
0.1128	9400	-
0.1140	9500	0.2117
0.1152	9600	-
0.1164	9700	-
0.1176	9800	-
0.1188	9900	-
0.1200	10000	0.1987
0.1212	10100	-
0.1224	10200	-
0.1236	10300	-
0.1248	10400	-
0.1260	10500	0.2011
0.1272	10600	-
0.1284	10700	-
0.1296	10800	-
0.1308	10900	-
0.1320	11000	0.1775
0.1332	11100	-
0.1344	11200	-
0.1356	11300	-
0.1368	11400	-
0.1380	11500	0.2048
0.1392	11600	-
0.1404	11700	-
0.1416	11800	-
0.1428	11900	-
0.1440	12000	0.2064
0.1452	12100	-
0.1464	12200	-
0.1476	12300	-
0.1488	12400	-
0.1500	12500	0.1883
0.1512	12600	-
0.1524	12700	-
0.1536	12800	-
0.1548	12900	-
0.1560	13000	0.2084
0.1572	13100	-
0.1584	13200	-
0.1596	13300	-
0.1608	13400	-
0.1620	13500	0.2077
0.1632	13600	-
0.1644	13700	-
0.1656	13800	-
0.1668	13900	-
0.1680	14000	0.1866
0.1692	14100	-
0.1704	14200	-
0.1716	14300	-
0.1728	14400	-
0.1740	14500	0.1859
0.1752	14600	-
0.1764	14700	-
0.1776	14800	-
0.1788	14900	-
0.1800	15000	0.1735
0.1812	15100	-
0.1824	15200	-
0.1836	15300	-
0.1848	15400	-
0.1860	15500	0.171
0.1872	15600	-
0.1884	15700	-
0.1896	15800	-
0.1908	15900	-
0.1920	16000	0.1465
0.1932	16100	-
0.1944	16200	-
0.1956	16300	-
0.1968	16400	-
0.1980	16500	0.1921
0.1992	16600	-
0.2004	16700	-
0.2016	16800	-
0.2028	16900	-
0.2040	17000	0.1669
0.2052	17100	-
0.2064	17200	-
0.2076	17300	-
0.2088	17400	-
0.2100	17500	0.1656
0.2112	17600	-
0.2124	17700	-
0.2136	17800	-
0.2148	17900	-
0.2160	18000	0.1952
0.2172	18100	-
0.2184	18200	-
0.2196	18300	-
0.2208	18400	-
0.2220	18500	0.1658
0.2232	18600	-
0.2244	18700	-
0.2256	18800	-
0.2268	18900	-
0.2280	19000	0.1774
0.2292	19100	-
0.2304	19200	-
0.2316	19300	-
0.2328	19400	-
0.2340	19500	0.1802
0.2352	19600	-
0.2364	19700	-
0.2376	19800	-
0.2388	19900	-
0.2400	20000	0.1724
0.2412	20100	-
0.2424	20200	-
0.2436	20300	-
0.2448	20400	-
0.2460	20500	0.1653
0.2472	20600	-
0.2484	20700	-
0.2496	20800	-
0.2508	20900	-
0.2520	21000	0.1484
0.2532	21100	-
0.2544	21200	-
0.2556	21300	-
0.2568	21400	-
0.2580	21500	0.1544
0.2592	21600	-
0.2604	21700	-
0.2616	21800	-
0.2628	21900	-
0.2640	22000	0.174
0.2652	22100	-
0.2664	22200	-
0.2676	22300	-
0.2688	22400	-
0.2700	22500	0.1488
0.2712	22600	-
0.2724	22700	-
0.2736	22800	-
0.2748	22900	-
0.2760	23000	0.1696
0.2772	23100	-
0.2784	23200	-
0.2796	23300	-
0.2808	23400	-
0.2820	23500	0.1468
0.2832	23600	-
0.2844	23700	-
0.2856	23800	-
0.2868	23900	-
0.2880	24000	0.1738
0.2892	24100	-
0.2904	24200	-
0.2916	24300	-
0.2928	24400	-
0.2940	24500	0.1667
0.2952	24600	-
0.2964	24700	-
0.2976	24800	-
0.2988	24900	-
0.3000	25000	0.1562
0.3012	25100	-
0.3024	25200	-
0.3036	25300	-
0.3048	25400	-
0.3060	25500	0.1628
0.3072	25600	-
0.3084	25700	-
0.3096	25800	-
0.3108	25900	-
0.3120	26000	0.1392
0.3132	26100	-
0.3144	26200	-

Framework Versions

• Python: 3.12.10
• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.8.1
• Datasets: 3.6.0
• Tokenizers: 0.21.1

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}
}