x1saint/gte-multi-triplet-v2

SentenceTransformer based on Alibaba-NLP/gte-multilingual-base

This is a sentence-transformers model finetuned from Alibaba-NLP/gte-multilingual-base on the all-nli-tr 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: Alibaba-NLP/gte-multilingual-base
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • all-nli-tr
• Language: 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': 8192, 'do_lower_case': False}) with Transformer model: NewModel 
  (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): 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("x1saint/gte-multi-triplet-v2")
# Run inference
sentences = [
    'Ve gerçekten, baba haklıydı, oğlu zaten her şeyi tecrübe etmişti, her şeyi denedi ve daha az ilgileniyordu.',
    'Oğlu her şeye olan ilgisini kaybediyordu.',
    'Baba oğlunun tecrübe için hala çok şey olduğunu biliyordu.',
]
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

• Dataset: all-nli-dev
• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9438

Training Details

Training Dataset

all-nli-tr

• Dataset: all-nli-tr at daeabfb
• Size: 482,091 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 5 tokens mean: 28.16 tokens max: 151 tokens	min: 5 tokens mean: 15.14 tokens max: 49 tokens	min: 4 tokens mean: 14.33 tokens max: 55 tokens

• Samples:

  anchor	positive	negative
  Mevsim boyunca ve sanırım senin seviyendeyken onları bir sonraki seviyeye düşürürsün. Eğer ebeveyn takımını çağırmaya karar verirlerse Braves üçlü A'dan birini çağırmaya karar verirlerse çifte bir adam onun yerine geçmeye gider ve bekar bir adam gelir.	Eğer insanlar hatırlarsa, bir sonraki seviyeye düşersin.	Hiçbir şeyi hatırlamazlar.
  Numaramızdan biri talimatlarınızı birazdan yerine getirecektir.	Ekibimin bir üyesi emirlerinizi büyük bir hassasiyetle yerine getirecektir.	Şu anda boş kimsek yok, bu yüzden sen de harekete geçmelisin.
  Bunu nereden biliyorsun? Bütün bunlar yine onların bilgileri.	Bu bilgi onlara ait.	Hiçbir bilgileri yok.

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

all-nli-tr

• Dataset: all-nli-tr at daeabfb
• Size: 6,567 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 3 tokens mean: 26.66 tokens max: 121 tokens	min: 5 tokens mean: 14.98 tokens max: 49 tokens	min: 4 tokens mean: 14.4 tokens max: 37 tokens

• Samples:

  anchor	positive	negative
  Bilemiyorum. Onunla ilgili karışık duygularım var. Bazen ondan hoşlanıyorum ama aynı zamanda birisinin onu dövmesini görmeyi seviyorum.	Çoğunlukla ondan hoşlanıyorum, ama yine de birinin onu dövdüğünü görmekten zevk alıyorum.	O benim favorim ve kimsenin onu yendiğini görmek istemiyorum.
  Sen ve arkadaşların burada hoş karşılanmaz, Severn söyledi.	Severn orada insanların hoş karşılanmadığını söyledi.	Severn orada insanların her zaman hoş karşılanacağını söyledi.
  Gecenin en aşağısı ne olduğundan emin değilim.	Dün gece ne kadar soğuk oldu bilmiyorum.	Dün gece hava 37 dereceydi.

• 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: 64
• per_device_eval_batch_size: 64
• gradient_accumulation_steps: 4
• num_train_epochs: 10
• warmup_ratio: 0.1
• bf16: True
• dataloader_num_workers: 4

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: 4
• 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: 1.0
• 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: True
• fp16: False
• 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: 4
• 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: 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
• 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: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	all-nli-dev_cosine_accuracy
0.2655	500	3.0729	0.4237	0.9229
0.5310	1000	2.2154	0.3830	0.9257
0.7965	1500	1.9267	0.3517	0.9319
1.0616	2000	1.7078	0.3424	0.9354
1.3271	2500	1.4602	0.3362	0.9368
1.5926	3000	1.3925	0.3290	0.9379
1.8581	3500	1.3124	0.3116	0.9417
2.1232	4000	1.1537	0.3154	0.9382
2.3887	4500	1.0209	0.3205	0.9412
2.6542	5000	0.9897	0.3065	0.9441
2.9197	5500	0.9611	0.3025	0.9420
3.1848	6000	0.8276	0.3162	0.9446
3.4503	6500	0.7779	0.3101	0.9408
3.7158	7000	0.7738	0.3110	0.9426
3.9813	7500	0.7641	0.3056	0.9434
4.2464	8000	0.6338	0.3152	0.9429
4.5119	8500	0.6397	0.3133	0.9421
4.7774	9000	0.6207	0.3160	0.9420
5.0425	9500	0.6044	0.3156	0.9408
5.3080	10000	0.5305	0.3205	0.9449
5.5735	10500	0.5377	0.3124	0.9450
5.8390	11000	0.5311	0.3168	0.9443
6.1041	11500	0.5017	0.3250	0.9435
6.3696	12000	0.46	0.3213	0.9429
6.6351	12500	0.4679	0.3212	0.9443
6.9006	13000	0.4692	0.3221	0.9434
7.1657	13500	0.4285	0.3231	0.9446
7.4312	14000	0.4161	0.3265	0.9456
7.6967	14500	0.409	0.3240	0.9456
7.9622	15000	0.4127	0.3250	0.9444
8.2273	15500	0.3843	0.3290	0.9447
8.4928	16000	0.3755	0.3259	0.9438
8.7583	16500	0.3786	0.3328	0.9438
9.0234	17000	0.3702	0.3284	0.9453
9.2889	17500	0.3525	0.3326	0.9444
9.5544	18000	0.3589	0.3320	0.9443
9.8199	18500	0.3483	0.3314	0.9438

Framework Versions

• Python: 3.11.11
• Sentence Transformers: 3.4.1
• Transformers: 4.48.3
• PyTorch: 2.5.1+cu124
• Accelerate: 1.3.0
• Datasets: 3.3.0
• Tokenizers: 0.21.0

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}
}