SentenceTransformer based on intfloat/multilingual-e5-base

This is a sentence-transformers model finetuned from intfloat/multilingual-e5-base. 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: intfloat/multilingual-e5-base
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, 'architecture': 'XLMRobertaModel'})
  (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("Funghang/e5-nepali-qa-ir")
# Run inference
queries = [
    "विवाहित व्यक्तिको हकमा साधारण राहदानीको लागि आवेदन कहाँ दिने ?",
]
documents = [
    'ताप्लेजुङ जिल्ला प्रशासन कार्यालयको सम्पर्क नम्बरहरू: नागरिकता फाँट : ०२४-४६०२७०, आर्थिक प्रशासन फाँट : ०२४-४६०५६६, स्थानीय प्रशासन फाँट : ०२४-४६०१९१ हुन् ।',
    'जिल्ला प्रशासन कार्यालय ओखलढुंगाको आधिकारिक इमेल अभिलेख प्रयोजनका लागिः avilekh.daookhaldhunga@gmail.com हो ।',
    'विवाहित व्यक्तिले आफ्नो स्थायी बसोबास भएको जिल्ला प्रशासन कार्यालय वा नागरिकता रहेको जिल्लास्थित जिल्ला प्रशासन कार्यालयमा आवश्यक कागजात सहित साधारण राहदानीको लागि निवेदन दिन सकिन्छ ।',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 768] [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.9078, 0.1416, 0.1064]])

Training Details

Training Dataset

Unnamed Dataset

Size: 3,840 training samples
Columns: query and positive
Approximate statistics based on the first 1000 samples:
query positive
type string string
details
min: 5 tokens
mean: 16.31 tokens
max: 65 tokens

min: 14 tokens
mean: 40.99 tokens
max: 338 tokens

	query	positive
type	string	string
details	min: 5 tokens mean: 16.31 tokens max: 65 tokens	min: 14 tokens mean: 40.99 tokens max: 338 tokens

Samples:

query	positive
`राहदानी कार्यालय संखुवासभाको सम्पर्क नम्बर के हो ?`	`राहदानी कार्यालय संखुवासभाको सम्पर्क नम्बरहरू ०२९५६०१३४, ०२९५६०१३३, ०२९५६०५३३ हुन् ।`
`राहदानीको लागि दर्ता केन्द्र कसरी छनोट गर्ने ?`	राहदानी बनाउन दर्ता केन्द्र छनोट गर्दा तपाईंले अनलाइन फाराम भर्दा नै आफ्नो लागि उपयुक्त स्थान चयन गर्नुपर्छ । नेपालमा रहेका जिल्ला वा इलाका प्रशासन कार्यालय वा राहदानी विभाग दर्ता केन्द्रका रूपमा छनोट गर्न सकिन्छ । विदेशमा रहेका नागरिकहरूले भने सम्बन्धित दूतावास वा कन्सुलेटलाई दर्ता केन्द्रका रूपमा छनोट गर्नुपर्छ । साथै, फाराम भर्दा Appointment लिनु पनि अनिवार्य हुन्छ जसले तपाईंलाई सुविधाजनक मिति र समय दिन्छ । यसरी छनोट गरिएको दर्ता केन्द्रमा नै तपाईंले सम्पूर्ण प्रक्रिया पूरा गर्नुपर्छ ।
`म संग राहदानी विभाग काठमाडौंबाट जारी भएको पुरानो हस्तलिखित राहदानी छ भने के मैले पुन राहदानी विभागबाट एमआरपि को लागि आवेदन दिन जिल्लाको सिफारिस ल्याउनु पर्छ ?`	`यदि तपाईंसँग राहदानी विभाग काठमाडौंबाट जारी भएको पुरानो हस्तलिखित राहदानी छ भने पनि पुन राहदानी विभागबाट एमआरपि को लागि आवेदन दिन जिल्लाको सिफारिस ल्याउनु पर्छ ।`

Loss: MultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

Size: 820 evaluation samples
Columns: query and positive
Approximate statistics based on the first 820 samples:
query positive
type string string
details
min: 7 tokens
mean: 17.08 tokens
max: 53 tokens

min: 14 tokens
mean: 45.07 tokens
max: 175 tokens

	query	positive
type	string	string
details	min: 7 tokens mean: 17.08 tokens max: 53 tokens	min: 14 tokens mean: 45.07 tokens max: 175 tokens

Samples:

query	positive
`के राहदानी बनाउन फोटोको आवश्यकता पर्छ ?`	`राहदानी बनाउन फोटोको आवश्यकता हुँदैन, किनभने फोटो खिच्ने काम कार्यालयमै हुन्छ ।`
`रुकुम पश्चिमको राहदानी कार्यालयसँग सम्पर्क गर्न कुन नम्बर प्रयोग गर्ने ?`	`रुकुम पश्चिमको राहदानी कार्यालयका लागि ०८८५३००४०, ०८८५३००९०, ०८८५३००९० नम्बर प्रयोग गर्न सकिन्छ ।`
`अफिसियल वा कूटनीतिक राहदानी भएकाले साधारण राहदानी बनाउँदा कुन विकल्प छान्ने ?`	`अफिसियल र कूटनीतिक राहदानी भएकाले साधारण राहदानी बनाउँदा आफ्नो पुरानो नम्बर बेवास्ता गरी उपयुक्त विकल्प छान्नु पर्छ।`

Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
learning_rate: 1e-05
num_train_epochs: 150
warmup_ratio: 0.1
fp16: True
load_best_model_at_end: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 8
per_device_eval_batch_size: 8
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: 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: 150
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: 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}
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_fused
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
hub_revision: None
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
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
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss
0.2083	100	0.9835	0.8339
0.4167	200	0.8657	0.6287
0.625	300	0.6003	0.3697
0.8333	400	0.3603	0.1957
1.0417	500	0.1578	0.1272
1.25	600	0.1	0.0974
1.4583	700	0.0619	0.0812
1.6667	800	0.0416	0.0751
1.875	900	0.0369	0.0714
2.0833	1000	0.0295	0.0676
2.2917	1100	0.0259	0.0641
2.5	1200	0.0168	0.0620
2.7083	1300	0.0514	0.0612
2.9167	1400	0.0294	0.0635
3.125	1500	0.0137	0.0605
3.3333	1600	0.013	0.0598
3.5417	1700	0.0193	0.0618
3.75	1800	0.0093	0.0600
3.9583	1900	0.0209	0.0623
4.1667	2000	0.0075	0.0654
4.375	2100	0.0152	0.0632
4.5833	2200	0.0205	0.0647
4.7917	2300	0.0062	0.0630
5.0	2400	0.0188	0.0616
5.2083	2500	0.0085	0.0596
5.4167	2600	0.0119	0.0605
5.625	2700	0.0087	0.0619
5.8333	2800	0.0115	0.0666
6.0417	2900	0.0203	0.0648
6.25	3000	0.0114	0.0644
6.4583	3100	0.0123	0.0650
6.6667	3200	0.0121	0.0609
6.875	3300	0.0044	0.0610
7.0833	3400	0.0061	0.0683
7.2917	3500	0.0151	0.0645

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.12.11
Sentence Transformers: 5.1.0
Transformers: 4.55.4
PyTorch: 2.8.0+cu126
Accelerate: 1.10.1
Datasets: 4.0.0
Tokenizers: 0.21.4

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}
}

Citation

@article{begha2026nepali,
  title={Nepali Passport Question Answering: A Low-Resource Dataset for Public Service Applications},
  author={Begha, Funghang Limbu and Acharya, Praveen and Bal, Bal Krishna},
  journal={arXiv preprint arXiv:2603.13320},
  year={2026}
}

E5 Model

@article{wang2024multilingual,
  title={Multilingual e5 text embeddings: A technical report},
  author={Wang, Liang and Yang, Nan and Huang, Xiaolong and Yang, Linjun and Majumder, Rangan and Wei, Furu},
  journal={arXiv preprint arXiv:2402.05672},
  year={2024}
}

Downloads last month: 60

Safetensors

Model size

0.3B params

Tensor type

F32

Model tree for Funghang/e5-nepali-qa-ir

Base model

intfloat/multilingual-e5-base

Finetuned

(119)

this model

Papers for Funghang/e5-nepali-qa-ir

Nepali Passport Question Answering: A Low-Resource Dataset for Public Service Applications

Paper • 2603.13320 • Published Mar 4