Intellexus/Bi-Tib-mbert-v1

SentenceTransformer based on Intellexus/mbert-tibetan-continual-wylie-final

This is a sentence-transformers model finetuned from Intellexus/mbert-tibetan-continual-wylie-final. 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: Intellexus/mbert-tibetan-continual-wylie-final
• Maximum Sequence Length: 512 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': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    "dam tshig nyams pa'i nyes pa ni/ 'dod pa'i phyogs mi 'grub cing / mi 'dod pa'i phyogs rnams thob pa ste/",
    "dam tshig dang ni mi ldan na// bsgrubs kyang 'grub par mi 'gyur te//\nrgyu med pa yi 'bras bu bzhin// tshe yi  dus byas dmyal bar 'gro//\n",
    "lha dang lha mo ji lta bas// bdud rtsi'i  bum pas dbang bskur ba//\nchu'i dgongs pa  ye shes lnga'i// rtags su  sku lnga rdzogs pa'o//\n",
]
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

Semantic Similarity

• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.835
spearman_cosine	0.854

Training Details

Training Dataset

Unnamed Dataset

• Size: 2,500 training samples
• Columns: text1, text2, and label
• Approximate statistics based on the first 1000 samples:

  	text1	text2	label
  type	string	string	float
  details	min: 6 tokens mean: 19.74 tokens max: 67 tokens	min: 5 tokens mean: 22.11 tokens max: 83 tokens	min: 0.02 mean: 0.51 max: 1.0

• Samples:

  text1	text2	label
  'on pa rnams kyang rna bas sgra thos p	'on pa rnams rna bas sgra thes par bya'o snyam pa dang / smyon pa rnams dran pa thob par	0.5
  com ldan 'das de bzhin gshegs pa dgra bc	mkhas pa yongs su gzung bar 'dod pa'i byang chub sems dpa' sems dpa' chen	0.229
  pa /sems can thams cad	ng / snying rje'i sems dang ldan pa	0.3335

• Loss: CoSENTLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 150 evaluation samples
• Columns: text1, text2, and label
• Approximate statistics based on the first 150 samples:

  	text1	text2	label
  type	string	string	float
  details	min: 8 tokens mean: 32.74 tokens max: 126 tokens	min: 6 tokens mean: 32.12 tokens max: 121 tokens	min: 0.0 mean: 0.5 max: 1.0

• Samples:

  text1	text2	label
  khang ljon shing rgyal mtshan seng ge rta	khang bzangs ljong shing bram ze seng ge rta	0.5625
  rnam par thar pa'i sgo mtshan ma med pa/	yod ces bya bar yang dag par rjes su mi mthong ba/	0.375
  byang chub ni chos kyi dbyings kyi gnas kyis gnas pa'o// byang chub ni de bzhin nyid rjes su rtogs pa'o//	nges pa yod na mngon sum min// 'dra bar 'dzin pa rtog pa yin//	0.0

• Loss: CoSENTLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 32
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 7
• load_best_model_at_end: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 8
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• 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: 7
• 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: 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: 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}
• 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
• 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	spearman_cosine
1.0	2	56.9409	2.7480	0.8357
2.0	4	53.1489	2.7016	0.8412
3.0	6	52.3657	2.6812	0.8462
3.8421	7	89.1774	2.6767	0.8471
0.8101	4	96.7978	2.7350	0.8455
1.8101	8	94.8279	2.6985	0.8497
2.8101	12	93.583	2.6846	0.8540

Framework Versions

• Python: 3.12.11
• Sentence Transformers: 4.1.0
• Transformers: 4.50.0
• PyTorch: 2.5.1
• Accelerate: 1.7.0
• Datasets: 3.3.2
• 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",
}

CoSENTLoss

@online{kexuefm-8847,
    title={CoSENT: A more efficient sentence vector scheme than Sentence-BERT},
    author={Su Jianlin},
    year={2022},
    month={Jan},
    url={https://kexue.fm/archives/8847},
}