NetherQuartz/paraphrase-MiniLM-tokipona

SentenceTransformer based on sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2

This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2. It maps sentences & paragraphs to a 384-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/paraphrase-multilingual-MiniLM-L12-v2
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 384 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': 128, 'do_lower_case': False, 'architecture': 'BertModel'})
  (1): Pooling({'word_embedding_dimension': 384, '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("sentence_transformers_model_id")
# Run inference
sentences = [
    '我只想暖和一下。',
    'mi wile kama seli taso.',
    'tomo tawa sina li lon ni.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.7361, 0.2725],
#         [0.7361, 1.0000, 0.2417],
#         [0.2725, 0.2417, 1.0000]])

Evaluation

Metrics

Knowledge Distillation

• Dataset: eval_data
• Evaluated with MSEEvaluator

Metric	Value
negative_mse	-1.9607

Translation

• Dataset: eval_data
• Evaluated with TranslationEvaluator

Metric	Value
src2trg_accuracy	0.6918
trg2src_accuracy	0.6346
mean_accuracy	0.6632

Training Details

Training Dataset

Unnamed Dataset

• Size: 82,069 training samples
• Columns: natural, tok, and label
• Approximate statistics based on the first 1000 samples:

  	natural	tok	label
  type	string	string	list
  details	min: 4 tokens mean: 10.99 tokens max: 45 tokens	min: 4 tokens mean: 15.64 tokens max: 55 tokens	size: 384 elements

• Samples:

  natural	tok	label
  Я держу руку.	mi sewi e luka mi.	[-0.17412713170051575, 0.2601699233055115, 0.3189601004123688, 0.009355960413813591, -0.030796436592936516, ...]
  Я змарыўся ад працы.	tan pali mi la mi pilin lape.	[0.1258312165737152, 0.173202782869339, 0.16050441563129425, 0.2519824206829071, -0.035661786794662476, ...]
  Mi bolso necesita ser reparado.	poki mi li pakala.	[-0.22065182030200958, 0.3290186822414398, -0.006242208182811737, 0.18535998463630676, 0.3087056577205658, ...]

• Loss: MSELoss

Evaluation Dataset

Unnamed Dataset

• Size: 4,267 evaluation samples
• Columns: natural, tok, and label
• Approximate statistics based on the first 1000 samples:

  	natural	tok	label
  type	string	string	list
  details	min: 5 tokens mean: 11.0 tokens max: 44 tokens	min: 4 tokens mean: 15.4 tokens max: 66 tokens	size: 384 elements

• Samples:

  natural	tok	label
  Da quanto tempo sei/state in Germania?	tenpo pi suli seme la sina lon ma Tosi?	[0.43582403659820557, 0.4226286709308624, 0.06436676532030106, -0.38238099217414856, -0.13951840996742249, ...]
  Habesne difficultatem hac re?	ni li ike tawa sina anu seme?	[0.22038640081882477, 0.03845325857400894, 0.20817194879055023, 0.08335897326469421, -0.10346948355436325, ...]
  אני לא הולך להפסיד.	mi kama ala anpa.	[0.3058338761329651, 0.06292764097452164, 0.019105680286884308, -0.04162227734923363, -0.10258055478334427, ...]

• Loss: MSELoss

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• learning_rate: 2e-05
• num_train_epochs: 12
• warmup_ratio: 0.1
• fp16: True
• load_best_model_at_end: 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: 2e-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: 12
• 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: 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: True
• 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: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss	Validation Loss	eval_data_negative_mse	eval_data_mean_accuracy
0.0779	100	0.0257	-	-	-
0.1559	200	0.0235	-	-	-
0.2338	300	0.0221	-	-	-
0.3118	400	0.0217	-	-	-
0.3897	500	0.0209	-	-	-
0.4677	600	0.0201	-	-	-
0.5456	700	0.0192	-	-	-
0.6235	800	0.0186	-	-	-
0.7015	900	0.0176	-	-	-
0.7794	1000	0.0171	-	-	-
0.8574	1100	0.0166	-	-	-
0.9353	1200	0.0159	-	-	-
1.0133	1300	0.0154	-	-	-
1.0912	1400	0.015	-	-	-
1.1691	1500	0.0145	-	-	-
1.2471	1600	0.0143	-	-	-
1.3250	1700	0.014	-	-	-
1.4030	1800	0.0139	-	-	-
1.4809	1900	0.0136	-	-	-
1.5588	2000	0.0134	0.0122	-2.3609	0.5603
1.6368	2100	0.0133	-	-	-
1.7147	2200	0.0133	-	-	-
1.7927	2300	0.0132	-	-	-
1.8706	2400	0.0131	-	-	-
1.9486	2500	0.0131	-	-	-
2.0265	2600	0.0129	-	-	-
2.1044	2700	0.0127	-	-	-
2.1824	2800	0.0125	-	-	-
2.2603	2900	0.0125	-	-	-
2.3383	3000	0.0124	-	-	-
2.4162	3100	0.0123	-	-	-
2.4942	3200	0.0122	-	-	-
2.5721	3300	0.0121	-	-	-
2.6500	3400	0.0123	-	-	-
2.7280	3500	0.0122	-	-	-
2.8059	3600	0.0122	-	-	-
2.8839	3700	0.0121	-	-	-
2.9618	3800	0.0122	-	-	-
3.0398	3900	0.012	-	-	-
3.1177	4000	0.0119	0.0110	-2.1275	0.6289
3.1956	4100	0.0118	-	-	-
3.2736	4200	0.0118	-	-	-
3.3515	4300	0.0117	-	-	-
3.4295	4400	0.0117	-	-	-
3.5074	4500	0.0116	-	-	-
3.5853	4600	0.0116	-	-	-
3.6633	4700	0.0117	-	-	-
3.7412	4800	0.0117	-	-	-
3.8192	4900	0.0116	-	-	-
3.8971	5000	0.0117	-	-	-
3.9751	5100	0.0115	-	-	-
4.0530	5200	0.0115	-	-	-
4.1309	5300	0.0113	-	-	-
4.2089	5400	0.0113	-	-	-
4.2868	5500	0.0114	-	-	-
4.3648	5600	0.0114	-	-	-
4.4427	5700	0.0113	-	-	-
4.5207	5800	0.0112	-	-	-
4.5986	5900	0.0113	-	-	-
4.6765	6000	0.0113	0.0107	-2.0522	0.6478
4.7545	6100	0.0112	-	-	-
4.8324	6200	0.0112	-	-	-
4.9104	6300	0.0113	-	-	-
4.9883	6400	0.0113	-	-	-
5.0663	6500	0.011	-	-	-
5.1442	6600	0.011	-	-	-
5.2221	6700	0.011	-	-	-
5.3001	6800	0.0109	-	-	-
5.3780	6900	0.0111	-	-	-
5.4560	7000	0.0111	-	-	-
5.5339	7100	0.011	-	-	-
5.6118	7200	0.0109	-	-	-
5.6898	7300	0.011	-	-	-
5.7677	7400	0.011	-	-	-
5.8457	7500	0.0111	-	-	-
5.9236	7600	0.011	-	-	-
6.0016	7700	0.0112	-	-	-
6.0795	7800	0.0108	-	-	-
6.1574	7900	0.0108	-	-	-
6.2354	8000	0.0107	0.0105	-2.0098	0.6524
6.3133	8100	0.0108	-	-	-
6.3913	8200	0.0108	-	-	-
6.4692	8300	0.0108	-	-	-
6.5472	8400	0.0109	-	-	-
6.6251	8500	0.0108	-	-	-
6.7030	8600	0.0108	-	-	-
6.7810	8700	0.0108	-	-	-
6.8589	8800	0.0107	-	-	-
6.9369	8900	0.0109	-	-	-
7.0148	9000	0.0108	-	-	-
7.0928	9100	0.0106	-	-	-
7.1707	9200	0.0107	-	-	-
7.2486	9300	0.0106	-	-	-
7.3266	9400	0.0105	-	-	-
7.4045	9500	0.0105	-	-	-
7.4825	9600	0.0107	-	-	-
7.5604	9700	0.0107	-	-	-
7.6383	9800	0.0108	-	-	-
7.7163	9900	0.0107	-	-	-
7.7942	10000	0.0106	0.0103	-1.9857	0.6582
7.8722	10100	0.0106	-	-	-
7.9501	10200	0.0106	-	-	-
8.0281	10300	0.0106	-	-	-
8.1060	10400	0.0105	-	-	-
8.1839	10500	0.0103	-	-	-
8.2619	10600	0.0105	-	-	-
8.3398	10700	0.0105	-	-	-
8.4178	10800	0.0105	-	-	-
8.4957	10900	0.0105	-	-	-
8.5737	11000	0.0105	-	-	-
8.6516	11100	0.0105	-	-	-
8.7295	11200	0.0106	-	-	-
8.8075	11300	0.0104	-	-	-
8.8854	11400	0.0106	-	-	-
8.9634	11500	0.0106	-	-	-
9.0413	11600	0.0105	-	-	-
9.1193	11700	0.0103	-	-	-
9.1972	11800	0.0102	-	-	-
9.2751	11900	0.0105	-	-	-
9.3531	12000	0.0104	0.0103	-1.9715	0.6624
9.4310	12100	0.0104	-	-	-
9.5090	12200	0.0104	-	-	-
9.5869	12300	0.0105	-	-	-
9.6648	12400	0.0104	-	-	-
9.7428	12500	0.0103	-	-	-
9.8207	12600	0.0105	-	-	-
9.8987	12700	0.0105	-	-	-
9.9766	12800	0.0104	-	-	-
10.0546	12900	0.0103	-	-	-
10.1325	13000	0.0103	-	-	-
10.2104	13100	0.0102	-	-	-
10.2884	13200	0.0103	-	-	-
10.3663	13300	0.0105	-	-	-
10.4443	13400	0.0103	-	-	-
10.5222	13500	0.0104	-	-	-
10.6002	13600	0.0104	-	-	-
10.6781	13700	0.0103	-	-	-
10.7560	13800	0.0103	-	-	-
10.8340	13900	0.0103	-	-	-
10.9119	14000	0.0103	0.0102	-1.9607	0.6632
10.9899	14100	0.0102	-	-	-
11.0678	14200	0.0103	-	-	-
11.1458	14300	0.0103	-	-	-
11.2237	14400	0.0103	-	-	-
11.3016	14500	0.0102	-	-	-
11.3796	14600	0.0104	-	-	-
11.4575	14700	0.0103	-	-	-
11.5355	14800	0.0103	-	-	-
11.6134	14900	0.0103	-	-	-
11.6913	15000	0.0102	-	-	-
11.7693	15100	0.0102	-	-	-
11.8472	15200	0.0103	-	-	-
11.9252	15300	0.0102	-	-	-
12.0	15396	-	0.0102	-1.9607	0.6632

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.13.7
• Sentence Transformers: 5.3.0
• Transformers: 4.55.2
• PyTorch: 2.10.0+cu128
• Accelerate: 1.13.0
• Datasets: 4.7.0
• 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",
}

MSELoss

@inproceedings{reimers-2020-multilingual-sentence-bert,
    title = "Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2020",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/2004.09813",
}