metadata

base_model: sentence-transformers/all-MiniLM-L6-v2
datasets:
  - momo22/eng2nep
language:
  - en
  - ne
library_name: sentence-transformers
metrics:
  - negative_mse
  - src2trg_accuracy
  - trg2src_accuracy
  - mean_accuracy
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:1000
  - loss:MSELoss
  - dataset_size:5000
  - dataset_size:8000
widget:
  - source_sentence: |
      The aggressive semi-employed religion workshop of Razzak, (EFP).
    sentences:
      - |
        मा ग्रिटर भेट्टाउन सकेन वा GDM प्रयोगकर्ताले कार्यान्वयन गर्न सकेन
      - |
        रज्जाकको आक्रामक अर्द्धशतक धर्मशाला, (एएफपी)।
      - |
        त्यसैले मेरो विजयपछि म त्यस्तो अवस्था आउन दिनेछैन।
  - source_sentence: >
      The authority is being a constitutional body, it was also empowered by
      passing the bill from Parliament.
    sentences:
      - >
        अख्तियार संवैधानिक निकाय त हुँदै हो, त्यसमा पनि संसदबाटै विधेयक पास गरेर
        अख्तियारलाई अधिकारसम्पन्न पनि गराइएको थियो।
      - >
        म यहूदाका राजा सिदकियाहलाई र उसका मानिसहरूलाई तिनीहरूका शत्रुहरूकहाँ
        सुम्पिनेछु जसले तिनीहरूलाई मार्न चाहन्छन्। ती सेनाहरू यरूशलेमबाट गइसकेका
        भएता पनि म तिनीहरूलाई बाबेलका राजाको सेनाहरूकहाँ सुम्पिनेछु।
      - |
        – संकटकालको असर न्यायिक क्षेत्रमा मात्रै पर्दैन, समग्र मुलुकमै पर्छ।
  - source_sentence: >
      The two-day conference will participate in investors from China, India,
      Japan, the US, European countries, Britain and other countries, the
      Federation said.
    sentences:
      - |
        उनीहरूको जनजीविकाको आधार प्राकृतिक स्रोत रहेको छ।
      - >
        दुई दिनसम्म हुने सम्मेलनमा चीन, भारत, जापान, अमेरिका, युरोपियन देशहरू,
        बेलायत लगायत देशबाट लगानीकर्ताको सहभागिता गराउने महासंघले जानकारी दिएको
        छ
      - |
        नयाँ स्न्यापसट लिनका लागि यो बटन क्लिक गर्नुहोस् ।
  - source_sentence: >
      Mr Sankey issued a "confession" through his solicitor after Shields had
      been convicted but then withdrew it.
    sentences:
      - >
        श्री सान्कीले ढालहरू दोषी भएपछि आफ्नो समाधानकर्तामार्फत "स्वीकृति" जारी
        गर्नुभयो तर त्यसपछि यसलाई फिर्ता लिनुभयो।
      - >
        कृत्रिम रुपमा पेट्रोलियम पदार्थको मूल्य स्थिर राख्न अनुदान दिदै जाने हो
        भने नेपाली अर्थतन्त्र एकदिन धराशायी हुनेछ।
      - >
        ओली सरकारले "राष्ट्रियता-राष्ट्रवाद र" आर्थिक सम्ब्रिद्धि "-आर्थिक
        विकासलाई यसको प्राथमिकताको रूपमा घोषणा गरेको छ।
  - source_sentence: >
      We want to use this time to appeal to the American government to see if
      they can finally close this chapter.
    sentences:
      - |
        धेरैले घाउ पाए र ओछ्यानमा थिए।
      - >
        नाम यसको अन्तरराष्ट्रिय हलको अद्वितिय डिजाइनबाट स्पष्ट रूपमा प्राप्त
        हुन्छ, जुन शीर्षकनियम स्पेसबाट बनेको छ, जुन ठूलो गहिराइमा उच्च दबाब
        बुझ्न सक्षम छ।
      - >
        हामी अमेरिकी सरकारलाई अपील गर्न यसपटक प्रयोग गर्न चाहन्छौं कि उनीहरूले
        अन्त्यमा यो अध्याय बन्द गर्न सक्छन्।
model-index:
  - name: SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2
    results:
      - task:
          type: knowledge-distillation
          name: Knowledge Distillation
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: negative_mse
            value: -0.37439612206071615
            name: Negative Mse
      - task:
          type: translation
          name: Translation
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: src2trg_accuracy
            value: 0.0186
            name: Src2Trg Accuracy
          - type: trg2src_accuracy
            value: 0.00835
            name: Trg2Src Accuracy
          - type: mean_accuracy
            value: 0.013474999999999999
            name: Mean Accuracy

SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2 on the momo22/eng2nep dataset. It maps sentences & paragraphs to a 384-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: sentence-transformers/all-MiniLM-L6-v2
Maximum Sequence Length: 256 tokens
Output Dimensionality: 384 tokens
Similarity Function: Cosine Similarity
Training Dataset:
- momo22/eng2nep
Languages: en, ne

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': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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})
  (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("jangedoo/all-MiniLM-L6-v2-nepali")
# Run inference
sentences = [
    'We want to use this time to appeal to the American government to see if they can finally close this chapter.\n',
    'हामी अमेरिकी सरकारलाई अपील गर्न यसपटक प्रयोग गर्न चाहन्छौं कि उनीहरूले अन्त्यमा यो अध्याय बन्द गर्न सक्छन्।\n',
    'नाम यसको अन्तरराष्ट्रिय हलको अद्वितिय डिजाइनबाट स्पष्ट रूपमा प्राप्त हुन्छ, जुन शीर्षकनियम स्पेसबाट बनेको छ, जुन ठूलो गहिराइमा उच्च दबाब बुझ्न सक्षम छ।\n',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Knowledge Distillation

Evaluated with MSEEvaluator

Metric	Value
negative_mse	-0.3744

Translation

Evaluated with TranslationEvaluator

Metric	Value
src2trg_accuracy	0.0186
trg2src_accuracy	0.0083
mean_accuracy	0.0135

Training Details

Training Dataset

momo22/eng2nep

Dataset: momo22/eng2nep at 57da8d4
Size: 8,000 training samples
Columns: English, Nepali, and label
Approximate statistics based on the first 1000 samples:
English Nepali label
type string string list
details
min: 3 tokens
mean: 26.29 tokens
max: 130 tokens

min: 3 tokens
mean: 65.39 tokens
max: 256 tokens

size: 384 elements

	English	Nepali	label
type	string	string	list
details	min: 3 tokens mean: 26.29 tokens max: 130 tokens	min: 3 tokens mean: 65.39 tokens max: 256 tokens	size: 384 elements

Samples:

English	Nepali	label
`But with the origin of feudal practices in the Middle Ages, the practice of untouchability began, as well as discrimination against women.`	`तर मध्ययुगमा सामन्ती प्रथाको उद्भव भएसँगै जसरी छुवाछुत प्रथाको शुरुवात भयो, त्यसैगरी नारी प्रति पनि विभेद गरिन थालियो`	`[-0.05432726442813873, 0.029996933415532112, -0.008532932959496975, -0.035200122743844986, 0.008856767788529396, ...]`
`A Pandit was found on the way to Pokhara from Baglung.`	`वाग्लुङ्गबाट पोखरा आउँदा बाटोमा एकजना पण्डित भेटिए।`	`[-0.023763148114085197, 0.0959007516503334, -0.11197677254676819, 0.10978179425001144, -0.028137238696217537, ...]`
`He went on: "She ate a perfectly normal and healthy diet.`	`उनी गए: "उनले पूर्ण सामान्य र स्वस्थ आहार खाइन्।`	`[0.028130479156970978, 0.030386686325073242, -0.012276170775294304, 0.1316223442554474, -0.01928202621638775, ...]`

Loss: MSELoss

Evaluation Dataset

momo22/eng2nep

Dataset: momo22/eng2nep at 57da8d4
Size: 500 evaluation samples
Columns: English, Nepali, and label
Approximate statistics based on the first 1000 samples:
English Nepali label
type string string list
details
min: 4 tokens
mean: 26.71 tokens
max: 213 tokens

min: 3 tokens
mean: 64.1 tokens
max: 256 tokens

size: 384 elements

	English	Nepali	label
type	string	string	list
details	min: 4 tokens mean: 26.71 tokens max: 213 tokens	min: 3 tokens mean: 64.1 tokens max: 256 tokens	size: 384 elements

Samples:

English	Nepali	label
`Chapter 3`	`परिच्छेद–३`	`[-0.049459926784038544, 0.048675183206796646, 0.016583453863859177, 0.04876156523823738, -0.020754676312208176, ...]`
`The capability of MOF would be strengthened to enable it to efficiently play the lead role in donor coordination, and to secure support from all stakeholders in aid coordination activities.`	`दाताहरूको समन्वयमा नेतृत्वदायीको भूमिका निर्वाह प्रभावकारी ढंगले गर्न अर्थ मन्त्रालयको क्षमता सुदृढ गरिनेछ यसको लागि सबै सरोकारवालाबाट समर्थन प्राप्त गरिनेछ ।`	`[-0.06200315058231354, -0.016507938504219055, -0.029924314469099045, -0.052509162575006485, 0.07746178656816483, ...]`
`Polimatrix, Inc. is a system integrator and total solutions provider delivering radiation and nuclear protection and detection.`	`पोलिमाट्रिक्स, इन्कर्पोरेटिड प्रणाली इन्टिजर र कुल समाधान प्रदायक रेडियो र आणविक संरक्षण र पत्ता लगाउने प्रणाली इन्टिजर र कुल समाधान प्रदायक हो।`	`[-0.0446796678006649, 0.026428330689668655, -0.09837698936462402, -0.07765442878007889, -0.020364686846733093, ...]`

Loss: MSELoss

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 64
per_device_eval_batch_size: 64
learning_rate: 2e-05
num_train_epochs: 1
warmup_ratio: 0.1
bf16: True
push_to_hub: True
hub_model_id: jangedoo/all-MiniLM-L6-v2-nepali
push_to_hub_model_id: all-MiniLM-L6-v2-nepali

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: 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: 1
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: 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: True
resume_from_checkpoint: None
hub_model_id: jangedoo/all-MiniLM-L6-v2-nepali
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: all-MiniLM-L6-v2-nepali
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: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	mean_accuracy	negative_mse
0.4	50	0.0021	0.0019	0.0111	-0.3837
0.8	100	0.002	0.0019	0.0123	-0.3794
0.4	50	0.002	0.0019	0.0130	-0.3773
0.8	100	0.002	0.0019	0.0135	-0.3744

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.0.1
Transformers: 4.42.4
PyTorch: 2.3.1+cu121
Accelerate: 0.32.1
Datasets: 2.21.0
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",
}

MSELoss

@inproceedings{reimers-2020-multilingual-sentence-bert,
    title = "Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2020",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/2004.09813",
}