Arabic-SBERT-100K

This is a sentence-transformers model finetuned from aubmindlab/bert-base-arabertv02. 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. This model is trained on 100K samples filtered from the akhooli/arabic-triplets-1m-curated-sims-len dataset with 75K training and 25K validation. Trained for 5 epochs, with final training loss of 0.133 (using MatryoshkaLoss).

The rest of this file is auto generated.

========================================================================

Model Details

Model Description

Model Type: Sentence Transformer
Base model: aubmindlab/bert-base-arabertv02
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 tokens
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 = [
    'ما هو نوع الدهون الموجودة في الأفوكادو',
    'حوالي 15 في المائة من الدهون في الأفوكادو مشبعة ، مع كل كوب واحد من الأفوكادو المفروم يحتوي على 3.2 جرام من الدهون المشبعة ، وهو ما يمثل 16 في المائة من DV البالغ 20 جرامًا. تحتوي الأفوكادو في الغالب على دهون أحادية غير مشبعة ، مع 67 في المائة من إجمالي الدهون ، أو 14.7 جرامًا لكل كوب مفروم ، ويتكون من هذا النوع من الدهون.',
    'يمكن أن يؤدي ارتفاع مستوى الدهون الثلاثية ، وهي نوع من الدهون (الدهون) في الدم ، إلى زيادة خطر الإصابة بأمراض القلب ، ويمكن أن يؤدي توفير مستوى مرتفع من الدهون الثلاثية ، وهي نوع من الدهون (الدهون) في الدم ، إلى زيادة خطر الإصابة بأمراض القلب. مرض.',
]
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]

Training Details

Training Dataset

Unnamed Dataset

Size: 75,000 training samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative
type	string	string	string
details	min: 4 tokens mean: 12.88 tokens max: 58 tokens	min: 4 tokens mean: 13.74 tokens max: 126 tokens	min: 4 tokens mean: 13.38 tokens max: 146 tokens

Samples:

anchor	positive	negative
`هل تشاجر (سي إس لويس) و (جي آر آر تولكين) ؟ إن كان الأمر كذلك، فما هو السبب؟`	`هل صحيح أن (سي إس لويس) و (تولكين) تشاجرا؟`	`ما هي أفضل الكتب للدراسة في الجامعة؟`
`ما هي اعراض فقر الدم؟`	`ما هي اعراض الانيميا؟`	`كيف احضر كيكة العسل؟`
`من ستصوت له، دونالد ترامب أم هيلاري كلينتون؟`	`هل تؤيدون دونالد ترامب أم هيلاري كلينتون؟ لماذا؟`	`كيف أتغلب على إدمان المواد الإباحية؟`

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "MultipleNegativesRankingLoss",
    "matryoshka_dims": [
        768,
        512,
        256,
        128,
        64
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Evaluation Dataset

Unnamed Dataset

Size: 25,000 evaluation samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative
type	string	string	string
details	min: 4 tokens mean: 12.6 tokens max: 70 tokens	min: 4 tokens mean: 14.82 tokens max: 239 tokens	min: 4 tokens mean: 13.78 tokens max: 128 tokens

Samples:

anchor	positive	negative
`نعم , نعم , أو رأيت " تشيما بارا ديسو "`	`نعم، أو "تشيما بارا ديسو" كانت تلك التي شاهدتها`	`أنا لم أرى "تشيما بارا ديسو".`
`رجل وامرأة يجلسان على الشاطئ بينما تغرب الشمس`	`هناك رجل وامرأة يجلسان على الشاطئ`	`إنهم يشاهدون شروق الشمس`
`كيف أسيطر على غضبي؟`	`ما هي أفضل طريقة للسيطرة على الغضب؟`	`كيف أعرف إن كانت زوجتي تخونني؟`

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "MultipleNegativesRankingLoss",
    "matryoshka_dims": [
        768,
        512,
        256,
        128,
        64
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
learning_rate: 2e-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: 16
per_device_eval_batch_size: 16
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_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: 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
eval_on_start: False
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss
0.2133	500	1.4163	0.3134
0.4266	1000	0.3306	0.1912
0.6399	1500	0.2263	0.1527
0.8532	2000	0.1818	0.1297
1.0666	2500	0.1658	0.1167
1.2799	3000	0.1139	0.1040
1.4932	3500	0.0808	0.1018
1.7065	4000	0.0692	0.0959
1.9198	4500	0.058	0.0958
2.1331	5000	0.0653	0.0882
2.3464	5500	0.0503	0.0912
2.5597	6000	0.0338	0.0970
2.7730	6500	0.0363	0.0906
2.9863	7000	0.0375	0.0856
3.1997	7500	0.0401	0.0879
3.4130	8000	0.031	0.0848
3.6263	8500	0.0255	0.0938
3.8396	9000	0.0239	0.0858
4.0529	9500	0.0305	0.0840
4.2662	10000	0.0281	0.0833
4.4795	10500	0.0174	0.0840
4.6928	11000	0.0216	0.0882
4.9061	11500	0.022	0.0866

Framework Versions

Python: 3.10.13
Sentence Transformers: 3.0.1
Transformers: 4.42.3
PyTorch: 2.1.2
Accelerate: 0.32.1
Datasets: 2.20.0
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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning}, 
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

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}
}

Downloads last month: 41,405

Safetensors

Model size

0.1B params

Tensor type

F32

Model tree for akhooli/Arabic-SBERT-100K

Base model

aubmindlab/bert-base-arabertv02

Finetuned

(4017)

this model

Dataset used to train akhooli/Arabic-SBERT-100K

Papers for akhooli/Arabic-SBERT-100K