SentenceTransformer based on Davlan/bert-base-multilingual-cased-finetuned-yoruba

This is a sentence-transformers model finetuned from Davlan/bert-base-multilingual-cased-finetuned-yoruba. 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 Sources

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("adejumobi/bert-base-multilingual-cased-finetuned-yoruba-IR")
# Run inference
sentences = [
    'Kini o yẹ ki Ilu India ṣe lori ikọlu UI?',
    'Bawo ni India le dahun si ikọlu ẹru UI?',
    'Lẹhin gbogbo họọsi ti media media ti ṣẹda awọn ikọlu URI Wip, kii yoo jẹ ohun itiju fun India ti ko ba kọlu Pakistan?',
]
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

Metric Value
cosine_accuracy 0.865
dot_accuracy 0.135
manhattan_accuracy 0.868
euclidean_accuracy 0.868
max_accuracy 0.868

Training Details

Training Dataset

Unnamed Dataset

  • Size: 5,019 training samples
  • Columns: query, pos, and neg
  • Approximate statistics based on the first 1000 samples:
    query pos neg
    type string string string
    details
    • min: 5 tokens
    • mean: 24.62 tokens
    • max: 74 tokens
    • min: 6 tokens
    • mean: 24.14 tokens
    • max: 79 tokens
    • min: 4 tokens
    • mean: 25.71 tokens
    • max: 98 tokens
  • Samples:
    query pos neg
    Kini idi ti Ilu India ṣe a ko ni ọkan lori ijiroro oloselu kan bi ni AMẸRIKA? Kini idi ti a ko le ni ijiroro gbangba laarin awọn oloselu ni India bi ọkan ninu wa? Njẹ eniyan le da quo duro de India Pakistan ariyanjiyan?A ni aisan ati ti o ri eyi lojoojumọ ni olopo?
    Kini OnePlus Ọkan? Bawo ni OnePlus kan? Kini idi ti OnePlus Ọkan dara?
    Ṣe ọkan wa ṣe iṣakoso awọn ẹdun wa? Bawo ni ọlọgbọn ati awọn eniyan aṣeyọri ṣe ṣakoso awọn ẹdun wọn? Bawo ni MO ṣe le ṣakoso awọn ẹdun mi rere fun awọn eniyan ti Mo nifẹ ṣugbọn wọn ko bikita nipa mi?
  • Loss: TripletLoss with these parameters:
    {
        "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
        "triplet_margin": 5
    }
    

Evaluation Dataset

Unnamed Dataset

  • Size: 1,000 evaluation samples
  • Columns: query, pos, and neg
  • Approximate statistics based on the first 1000 samples:
    query pos neg
    type string string string
    details
    • min: 6 tokens
    • mean: 24.32 tokens
    • max: 94 tokens
    • min: 6 tokens
    • mean: 24.06 tokens
    • max: 115 tokens
    • min: 6 tokens
    • mean: 25.58 tokens
    • max: 121 tokens
  • Samples:
    query pos neg
    Bawo ni o jẹ ọjọ ebi? Bawo ni o jẹ ọsan Njẹ NEBM lueMo ṣẹlẹ lati wa awọn ifiweranṣẹ ti o sọ pe o jẹ iro ati pe ko ni itter
    Kini awọn ohun elo akọkọ ti kọnputa kan? Kini diẹ ninu awọn ẹya akọkọ ti kọnputa kan?Awọn iṣẹ wo ni wọn nṣe iranṣẹ? Kini awọn eto kọmputa?Kini awọn iṣẹ ti awọn eto kọnputa?
    Ṣe o le faffiti Artists fun sokiri Graffiti ni Rockdale County, GA? Ṣe o le fun awọn ojukokoro fun fun sokiri Graffiti ni Cockdale County, Georgia? Kini idi ti Graffiti jẹ arufin?
  • Loss: TripletLoss with these parameters:
    {
        "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
        "triplet_margin": 5
    }
    

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: steps
  • per_device_train_batch_size: 12
  • per_device_eval_batch_size: 3
  • learning_rate: 1e-05
  • num_train_epochs: 5
  • 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: 12
  • per_device_eval_batch_size: 3
  • per_gpu_train_batch_size: None
  • per_gpu_eval_batch_size: None
  • gradient_accumulation_steps: 1
  • eval_accumulation_steps: None
  • learning_rate: 1e-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: 5
  • 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
  • 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
  • batch_sampler: no_duplicates
  • multi_dataset_batch_sampler: proportional

Training Logs

Epoch Step Training Loss loss cosine_accuracy
0 0 - - 0.827
0.2387 100 4.247 3.6056 0.815
0.4773 200 3.3576 2.7548 0.809
0.7160 300 2.931 2.3805 0.843
0.9547 400 2.4476 2.1895 0.858
1.1933 500 2.5839 2.1148 0.854
1.4320 600 2.0645 2.0497 0.855
1.6706 700 1.8386 2.0328 0.847
1.9093 800 1.5527 1.9380 0.857
2.1480 900 1.7298 1.8999 0.861
2.3866 1000 1.4375 1.8744 0.855
2.6253 1100 1.1605 1.8761 0.861
2.8640 1200 1.0601 1.8658 0.862
3.1026 1300 1.1019 1.8181 0.861
3.3413 1400 1.052 1.8088 0.854
3.5800 1500 0.8807 1.7937 0.862
3.8186 1600 0.7877 1.7963 0.862
4.0573 1700 0.7613 1.7869 0.868
4.2959 1800 0.8018 1.7696 0.867
4.5346 1900 0.6717 1.7815 0.865
4.7733 2000 0.6603 1.7776 0.865

Framework Versions

  • Python: 3.10.13
  • Sentence Transformers: 3.0.1
  • Transformers: 4.41.2
  • PyTorch: 2.1.2
  • Accelerate: 0.31.0
  • Datasets: 2.19.2
  • Tokenizers: 0.19.1

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",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification}, 
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}
Downloads last month
13
Safetensors
Model size
178M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for adejumobi/bert-base-multilingual-cased-finetuned-yoruba-IR

Finetuned
(1)
this model

Dataset used to train adejumobi/bert-base-multilingual-cased-finetuned-yoruba-IR

Evaluation results