Add new SentenceTransformer model

c32d12a verified over 1 year ago

20.3 kB

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:23478
  - loss:ContrastiveLoss
base_model: denaya/indoSBERT-large
widget:
  - source_sentence: 'Pekerja anak Indonesia: Buku panduan 2022 (pr & pasca pandemi)'
    sentences:
      - Statistik Perusahaan Hak Pengusahaan Hutan 2010
      - Statistik Kriminal 2016
      - ' Buletin Statistik Perdagangan Luar Negeri Ekspor Menurut Kelompok Komoditi dan Negara, November 2020'
  - source_sentence: Jumlah pascr tradisional, pusat perbelanjaan, dan toko modern tahun 2019
    sentences:
      - Profil Pasar Tradisional, Pusat Perbelanjaan, dan Toko Modern 2019
      - Laporan Perekonomian Indonesia 2008
      - Profil Industri Mikro dan Kecil 2006
  - source_sentence: Survei biay ahidup (SBH) di Ternate tahun 2012
    sentences:
      - Laporan Bulanan Data Sosial Ekonomi Januari 2016
      - Keadaan Angkatan kerja di Indonesia Agustus 2009
      - Statistik Perdagangan Luar Negeri Indonesia Impor 2023 Buku I
  - source_sentence: Direktori perwsahaan air minum, listrik, dan gas di kota tahun 2009
    sentences:
      - Statistik Indonesia 1991
      - Direktori Perusahaan Air Minum Listrik dan Gas Kota 2009
      - Direktori Eksportir Indonesia 2015
  - source_sentence: Studi efisiensi industri manufaktr
    sentences:
      - Statistik Indonesia 2019
      - Statistik Potensi Desa Provinsi Maluku 2011
      - Klasifikasi Baku Komoditas Indonesia (KBKI) 2012 Buku 4
datasets:
  - yahyaabd/bps-publication-pos-neg-pairs
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - pearson_cosine
  - spearman_cosine
model-index:
  - name: SentenceTransformer based on denaya/indoSBERT-large
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: allstats semantic base v1 eval
          type: allstats-semantic-base-v1-eval
        metrics:
          - type: pearson_cosine
            value: 0.9658815836712943
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.7841756166101173
            name: Spearman Cosine
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: allstat semantic base v1 test
          type: allstat-semantic-base-v1-test
        metrics:
          - type: pearson_cosine
            value: 0.9592021090962591
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.7818288777895762
            name: Spearman Cosine

SentenceTransformer based on denaya/indoSBERT-large

This is a sentence-transformers model finetuned from denaya/indoSBERT-large on the bps-publication-pos-neg-pairs dataset. 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
Training Dataset:
- bps-publication-pos-neg-pairs

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("yahyaabd/allstats-semantic-base-v1-3")
# Run inference
sentences = [
    'Studi efisiensi industri manufaktr',
    'Statistik Potensi Desa Provinsi Maluku 2011',
    'Klasifikasi Baku Komoditas Indonesia (KBKI) 2012 Buku 4',
]
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

Semantic Similarity

Datasets: allstats-semantic-base-v1-eval and allstat-semantic-base-v1-test
Evaluated with EmbeddingSimilarityEvaluator

Metric	allstats-semantic-base-v1-eval	allstat-semantic-base-v1-test
pearson_cosine	0.9659	0.9592
spearman_cosine	0.7842	0.7818

Training Details

Training Dataset

bps-publication-pos-neg-pairs

Dataset: bps-publication-pos-neg-pairs at 46a5cb7
Size: 23,478 training samples
Columns: query, doc, and label
Approximate statistics based on the first 1000 samples:
query doc label
type string string int
details
min: 5 tokens
mean: 11.84 tokens
max: 24 tokens

min: 5 tokens
mean: 10.77 tokens
max: 28 tokens

0: ~72.40%
1: ~27.60%

	query	doc	label
type	string	string	int
details	min: 5 tokens mean: 11.84 tokens max: 24 tokens	min: 5 tokens mean: 10.77 tokens max: 28 tokens	0: ~72.40% 1: ~27.60%

Samples:

query	doc	label
`Direktori perusahaan perantara keuangan bukan koperasi tahun 2006 (SE)`	`Tinjauan Regional Berdasarkan PDRB Kabupaten/Kota 2018-2022, Buku 2 Pulau Jawa-Bali`	`0`
`Informasi lengkap tentang PPLS 2011`	`Indeks Harga Perdagangan Besar Indonesia tahun 2005`	`0`
`Data konversi GKG ke beras tahun 2012`	`Indikator Ekonomi Juli 2023`	`0`

Loss: ContrastiveLoss with these parameters:

{
    "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
    "margin": 0.5,
    "size_average": true
}

Evaluation Dataset

bps-publication-pos-neg-pairs

Dataset: bps-publication-pos-neg-pairs at 46a5cb7
Size: 5,031 evaluation samples
Columns: query, doc, and label
Approximate statistics based on the first 1000 samples:
query doc label
type string string int
details
min: 5 tokens
mean: 11.97 tokens
max: 24 tokens

min: 5 tokens
mean: 10.76 tokens
max: 32 tokens

0: ~72.70%
1: ~27.30%

	query	doc	label
type	string	string	int
details	min: 5 tokens mean: 11.97 tokens max: 24 tokens	min: 5 tokens mean: 10.76 tokens max: 32 tokens	0: ~72.70% 1: ~27.30%

Samples:

query	doc	label
`Informasi angka tanaman berkhasiat ogbat dan tanaman hias di tahun 2005`	`Tinjauan Regional Berdasarkan PDRB Kabupaten/Kota 2010-2013 - Buku 2 Pulau Jawa-Bali`	`0`
`Informasi lengkap statistik horsikultura tahun 2020`	`NERACA ENERGI INDONESIA 2017-2021`	`0`
`Statistik air bersih Indonesia periode 2014-2019`	`Profil Usaha Konstruksi Perorangan Provinsi Kalimantan Utara, 2022`	`0`

Loss: ContrastiveLoss with these parameters:

{
    "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
    "margin": 0.5,
    "size_average": true
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 64
per_device_eval_batch_size: 64
num_train_epochs: 8
warmup_ratio: 0.1
fp16: True
load_best_model_at_end: True
eval_on_start: 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: 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
torch_empty_cache_steps: None
learning_rate: 5e-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: 8
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
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
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
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: True
use_liger_kernel: False
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	allstats-semantic-base-v1-eval_spearman_cosine	allstat-semantic-base-v1-test_spearman_cosine
0	0	-	0.0053	0.7770	-
0.5450	200	0.0023	0.0005	0.7842	-
1.0899	400	0.0005	0.0002	0.7842	-
1.6349	600	0.0002	0.0002	0.7842	-
2.1798	800	0.0001	0.0001	0.7842	-
2.7248	1000	0.0001	0.0001	0.7842	-
3.2698	1200	0.0	0.0001	0.7842	-
3.8147	1400	0.0	0.0001	0.7842	-
4.3597	1600	0.0	0.0001	0.7842	-
4.9046	1800	0.0	0.0001	0.7842	-
5.4496	2000	0.0	0.0001	0.7842	-
5.9946	2200	0.0	0.0001	0.7842	-
6.5395	2400	0.0	0.0001	0.7842	-
7.0845	2600	0.0	0.0001	0.7842	-
7.6294	2800	0.0	0.0001	0.7842	-
-1	-1	-	-	-	0.7818

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.4.0
Transformers: 4.48.1
PyTorch: 2.5.1+cu124
Accelerate: 1.3.0
Datasets: 3.2.0
Tokenizers: 0.21.0

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",
}

ContrastiveLoss

@inproceedings{hadsell2006dimensionality,
    author={Hadsell, R. and Chopra, S. and LeCun, Y.},
    booktitle={2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)},
    title={Dimensionality Reduction by Learning an Invariant Mapping},
    year={2006},
    volume={2},
    number={},
    pages={1735-1742},
    doi={10.1109/CVPR.2006.100}
}