annisamukhri/indosbert-climate-faq

SentenceTransformer based on denaya/indoSBERT-large

This is a sentence-transformers model finetuned from denaya/indoSBERT-large. It maps sentences & paragraphs to a 256-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: denaya/indoSBERT-large
• Maximum Sequence Length: 256 tokens
• Output Dimensionality: 256 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': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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): Dense({'in_features': 1024, 'out_features': 256, 'bias': True, 'activation_function': 'torch.nn.modules.activation.Tanh'})
)

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 = [
    '1 usul beton',
    'concrete berasal latin concretus bentuk pasif sempurna concrescere concrescere berasal con crescere tumbuh',
    'pencapaian utama earth summit 1992 meliputi pembentukan unfccc kesepakatan konvensi perubahan iklim kesepakatan aktivitas tanah masyarakat adat menyebabkan degradasi lingkungan sesuai budaya konvensi keanekaragaman hayati dibuka ditandatangani deklarasi rio lingkungan pembangunan agenda 21 prinsipprinsip kehutanan disetujui',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 256]

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

Evaluation

Metrics

Information Retrieval

• Dataset: indoSBERT-large-eval
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.836
cosine_accuracy@3	0.9212
cosine_accuracy@5	0.94
cosine_accuracy@10	0.9624
cosine_precision@1	0.836
cosine_precision@3	0.3071
cosine_precision@5	0.188
cosine_precision@10	0.0962
cosine_recall@1	0.836
cosine_recall@3	0.9212
cosine_recall@5	0.94
cosine_recall@10	0.9624
cosine_ndcg@10	0.9018
cosine_mrr@10	0.8821
cosine_map@100	0.8832

Training Details

Training Dataset

Unnamed Dataset

• Size: 6,461 training samples
• Columns: question and answer
• Approximate statistics based on the first 1000 samples:

  	question	answer
  type	string	string
  details	min: 3 tokens mean: 9.18 tokens max: 30 tokens	min: 3 tokens mean: 39.65 tokens max: 256 tokens

• Samples:

  question	answer
  sektor industri dikaitkan konflik lingkungan	sektor industri dikaitkan konflik lingkungan pertambangan energi fosil biomassa pemanfaatan lahan pengelolaan air sektorsektor mencakup 67 konflik lingkungan terdokumentasi atlas keadilan lingkungan
  ilmu teknik lingkungan berbeda teknik lingkungan ilmu lingkungan	ilmu teknik lingkungan memiliki mata kuliah teknik lingkungan dibandingkan ilmu lingkungan mata kuliah mengikuti kurikulum teknik lingkungan kuliah mahasiswa teknik lingkungan memilih bidangbidang desain fasilitas penyimpanan nuklir bioreaktor bakteri kebijakan lingkungan mahasiswa teknik lingkungan berfokus pembangunan fasilitas pengolahan penilaian dampak lingkungan mitigasi polusi udara
  perusahaan manakah kali menemukan minyak nigeria	shellbp menemukan minyak nigeria oloibiri 1956

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 1,384 evaluation samples
• Columns: question and answer
• Approximate statistics based on the first 1000 samples:

  	question	answer
  type	string	string
  details	min: 3 tokens mean: 8.85 tokens max: 27 tokens	min: 3 tokens mean: 39.69 tokens max: 256 tokens

• Samples:

  question	answer
  dampak potensial perubahan iklim ketersediaan air somalia	proyeksi ketersediaan air somalia berdasarkan skenario emisi mempertimbangkan pertumbuhan populasi model peningkatan sejalan proyeksi curah hujan mempertimbangkan proyeksi pertumbuhan populasi ketersediaan air kapita berkurang setengahnya 2080 berdasarkan skenario emisi rcp26 rcp60 ketidakpastian seputar volume air tersedia diproyeksikan
  peran neeri rencana implementasi nasional nip pops	neeri memainkan peran organisasi mitra rencana implementasi nasional nip pop india berkontribusi upaya negara mengatasi polutan organik persisten
  perubahan iklim mempengaruhi pertanian connecticut	suhu hangat mengurangi hasil industri susu connecticut bernilai 70 juta sapi makan menghasilkan susu cuaca panas peternakan dirugikan harihari panas kekeringan banjir mengurangi hasil panen menunda tanggal tanam peternakan diuntungkan musim tanam efek pemupukan karbon dioksida

• Loss: MultipleNegativesRankingLoss with these parameters:

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

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
• 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: 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
• use_liger_kernel: False
• eval_use_gather_object: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	indoSBERT-large-eval_cosine_ndcg@10
0	0	-	-	0.6184
0.2475	100	0.3999	0.1869	0.7909
0.4950	200	0.1581	0.1060	0.8580
0.7426	300	0.1107	0.0884	0.8738
0.9901	400	0.1028	0.0822	0.8872
1.2376	500	0.0784	0.0694	0.8886
1.4851	600	0.015	0.0764	0.8891
1.7327	700	0.0052	0.0757	0.8921
1.9802	800	0.0061	0.0691	0.8914
2.2277	900	0.0051	0.0723	0.8943
2.4752	1000	0.0052	0.0709	0.8950
2.7228	1100	0.0013	0.0729	0.8968
2.9703	1200	0.001	0.0703	0.8984
3.2178	1300	0.0019	0.0649	0.9002
3.4653	1400	0.0007	0.0654	0.8989
3.7129	1500	0.0004	0.0668	0.8997
3.9604	1600	0.0005	0.0681	0.9002
4.2079	1700	0.0004	0.0676	0.9016
4.4554	1800	0.001	0.0666	0.9012
4.7030	1900	0.0003	0.0667	0.9012
4.9505	2000	0.0003	0.0670	0.9018

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.3.1
• Transformers: 4.45.2
• PyTorch: 2.5.1+cu121
• Accelerate: 1.1.1
• Datasets: 3.1.0
• Tokenizers: 0.20.3

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}
}