This is a sentence-transformers model finetuned from microsoft/mpnet-base on the all-nli-triplets-turkish dataset. 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: microsoft/mpnet-base
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- all-nli-triplets-turkish
Languages: en, tr

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: MPNetModel 
  (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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("mertcobanov/mpnet-base-all-nli-triplet-turkish-v4-dgx")
# Run inference
sentences = [
    'Böyle şeyler görmek ve eğer yapabileceğiniz en küçük bir şey varsa, bu yardımcı olur.',
    'Böyle bir şeyi gözlemlemek ve yapıp yapamayacağınızı bilmek için.',
    'Böyle bir şeyi görmek kötü, eğer yapabiliyorsanız buna hiç katkıda bulunmayın.',
]
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]

Evaluation

Metrics

Triplet

Datasets: all-nli-dev-turkish and all-nli-test-turkish
Evaluated with TripletEvaluator

Metric	all-nli-dev-turkish	all-nli-test-turkish
cosine_accuracy	0.7764	0.7741

Training Details

Training Dataset

all-nli-triplets-turkish

Dataset: all-nli-triplets-turkish at 13554fd
Size: 120,781 training samples
Columns: anchor_translated, positive_translated, and negative_translated

Approximate statistics based on the first 1000 samples:

	anchor_translated	positive_translated	negative_translated
type	string	string	string
details	min: 3 tokens mean: 11.77 tokens max: 40 tokens	min: 3 tokens mean: 11.1 tokens max: 46 tokens	min: 2 tokens mean: 12.41 tokens max: 44 tokens

Samples:

anchor_translated	positive_translated	negative_translated
`Bir kişi, bir atın üzerinde, bozulmuş bir uçağın üzerinden atlıyor.`	`Bir kişi dışarıda, bir atın üzerinde.`	`Bir kişi bir lokantada omlet siparişi veriyor.`
`Bir Küçük Lig takımı, bir oyuncunun bir üsse kayarak girmeye çalıştığı sırada onu yakalamaya çalışıyor.`	`Bir takım bir koşucuyu dışarı atmaya çalışıyor.`	`Bir takım Satürn'de beyzbol oynuyor.`
`Kadın beyaz giyiyor.`	`Beyaz bir ceket giymiş bir kadın bir tekerlekli sandalyeyi itiyor.`	`Siyah giyinmiş bir adam, siyah giyinmiş bir kadını kucaklıyor.`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim" 4}

Evaluation Dataset

all-nli-triplets-turkish

Dataset: all-nli-triplets-turkish at 13554fd
Size: 6,584 evaluation samples
Columns: anchor_translated, positive_translated, and negative_translated

Approximate statistics based on the first 1000 samples:

	anchor_translated	positive_translated	negative_translated
type	string	string	string
details	min: 2 tokens mean: 22.3 tokens max: 135 tokens	min: 1 tokens mean: 10.92 tokens max: 41 tokens	min: 3 tokens mean: 10.81 tokens max: 34 tokens

Samples:

anchor_translated	positive_translated	negative_translated
`Ayrıca, bu özel tüketim vergileri, diğer vergiler gibi, hükümetin ödeme zorunluluğunu sağlama yetkisini kullanarak belirlenir.`	`Hükümetin ödeme zorlaması, özel tüketim vergilerinin nasıl hesaplandığını belirler.`	`Özel tüketim vergileri genel kuralın bir istisnasıdır ve aslında GSYİH payına dayalı olarak belirlenir.`
`Gri bir sweatshirt giymiş bir sanatçı, canlı renklerde bir kasaba tablosu üzerinde çalışıyor.`	`Bir ressam gri giysiler içinde bir kasabanın resmini yapıyor.`	`Bir kişi bir beyzbol sopası tutuyor ve gelen bir atış için planda bekliyor.`
`İmkansız.`	`Yapılamaz.`	`Tamamen mümkün.`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim" 4}

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: 10
warmup_ratio: 0.1
fp16: True
batch_sampler: no_duplicates

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: 10
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: 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: False
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: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	all-nli-dev-turkish_cosine_accuracy	all-nli-test-turkish_cosine_accuracy
0	0	-	-	0.5729	-
0.2119	100	6.6103	4.5154	0.6970	-
0.4237	200	5.1602	3.7328	0.7195	-
0.6356	300	4.4533	3.3389	0.7372	-
0.8475	400	3.4465	3.6044	0.7187	-
1.0572	500	2.6977	3.3043	0.7418	-
1.2691	600	3.8142	3.2066	0.7512	-
1.4809	700	3.4333	3.0716	0.7508	-
1.6928	800	3.1488	2.9590	0.7553	-
1.9047	900	1.8677	3.2416	0.7442	-
2.1144	1000	2.2034	2.9323	0.7634	-
2.3263	1100	2.9834	2.9406	0.7669	-
2.5381	1200	2.6785	2.8607	0.7672	-
2.75	1300	2.5096	2.8939	0.7684	-
2.9619	1400	0.876	3.2539	0.7416	-
3.1716	1500	2.3355	2.7503	0.7758	-
3.3835	1600	2.4666	2.7920	0.7707	-
3.5953	1700	2.2691	2.7860	0.7729	-
3.8072	1800	1.8024	2.9899	0.7571	-
4.0169	1900	0.6443	3.0993	0.7456	-
4.2288	2000	2.3976	2.7792	0.7811	-
4.4407	2100	2.1145	2.7968	0.7728	-
4.6525	2200	1.9788	2.7243	0.7751	-
4.8644	2300	1.1676	2.9885	0.7567	-
5.0742	2400	1.0009	2.7374	0.7767	-
5.2860	2500	2.1276	2.7822	0.7767	-
5.4979	2600	1.8459	2.7822	0.7760	-
5.7097	2700	1.7659	2.7322	0.7766	-
5.9216	2800	0.5916	3.0191	0.7596	-
6.1314	2900	1.3908	2.6973	0.7772	-
6.3432	3000	1.9257	2.7585	0.7763	-
6.5551	3100	1.6558	2.7350	0.7760	-
6.7669	3200	1.5368	2.7903	0.7722	-
6.9788	3300	0.1968	3.0849	0.7479	-
7.1886	3400	1.8044	2.6626	0.7825	-
7.4004	3500	1.7048	2.7380	0.7790	-
7.6123	3600	1.5666	2.7250	0.7796	-
7.8242	3700	1.0954	2.9620	0.7629	-
8.0339	3800	0.487	2.8900	0.7641	-
8.2458	3900	1.8398	2.7186	0.7796	-
8.4576	4000	1.5659	2.7259	0.7778	-
8.6695	4100	1.4825	2.7007	0.7760	-
8.8814	4200	0.7019	2.9050	0.7675	-
9.0911	4300	0.9278	2.7606	0.7731	-
9.3030	4400	1.766	2.6978	0.7787	-
9.5148	4500	1.4699	2.7114	0.7801	-
9.7267	4600	1.4647	2.7096	0.7799	-
9.9386	4700	0.3321	2.7418	0.7764	-
9.9809	4720	-	-	-	0.7741

Framework Versions

Python: 3.10.14
Sentence Transformers: 3.3.1
Transformers: 4.46.3
PyTorch: 2.4.0
Accelerate: 0.27.2
Datasets: 3.1.0
Tokenizers: 0.20.3

Citation

BibTeX

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{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}
}