yahyaabd/allstats-semantic-search-model-v1-3

SentenceTransformer based on sentence-transformers/paraphrase-multilingual-mpnet-base-v2

This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-multilingual-mpnet-base-v2 on the allstats-semantic-search-synthetic-dataset-v1 dataset. 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: sentence-transformers/paraphrase-multilingual-mpnet-base-v2
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • allstats-semantic-search-synthetic-dataset-v1

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': 128, 'do_lower_case': False}) with Transformer model: 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("yahyaabd/allstats-semantic-search-model-v1-3")
# Run inference
sentences = [
    'perubahan nilai tukar petani bulan mei 2017',
    'Perkembangan Nilai Tukar Petani Mei 2017',
    'Statistik Restoran/Rumah Makan Tahun 2014',
]
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]

Evaluation

Metrics

Semantic Similarity

• Datasets: allstats-semantic-search-v1-3-dev and allstat-semantic-search-v1-3-test
• Evaluated with EmbeddingSimilarityEvaluator

Metric	allstats-semantic-search-v1-3-dev	allstat-semantic-search-v1-3-test
pearson_cosine	0.9959	0.9961
spearman_cosine	0.9641	0.9648

Training Details

Training Dataset

allstats-semantic-search-synthetic-dataset-v1

• Dataset: allstats-semantic-search-synthetic-dataset-v1 at b13c0a7
• Size: 212,940 training samples
• Columns: query, doc, and label
• Approximate statistics based on the first 1000 samples:

  	query	doc	label
  type	string	string	float
  details	min: 5 tokens mean: 11.46 tokens max: 34 tokens	min: 5 tokens mean: 14.47 tokens max: 54 tokens	min: 0.0 mean: 0.5 max: 1.05

• Samples:

  query	doc	label
  aDta industri besar dan sedang Indonesia 2008	Statistik Industri Besar dan Sedang Indonesia 2008	0.9
  profil bisnis konstruksi individu jawa barat 2022	Statistik Industri Manufaktur Indonesia 2015 - Bahan Baku	0.15
  data statistik ekonomi indonesia	Nilai Tukar Valuta Asing di Indonesia 2014	0.08

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Evaluation Dataset

allstats-semantic-search-synthetic-dataset-v1

• Dataset: allstats-semantic-search-synthetic-dataset-v1 at b13c0a7
• Size: 26,618 evaluation samples
• Columns: query, doc, and label
• Approximate statistics based on the first 1000 samples:

  	query	doc	label
  type	string	string	float
  details	min: 5 tokens mean: 11.38 tokens max: 34 tokens	min: 4 tokens mean: 14.63 tokens max: 55 tokens	min: 0.0 mean: 0.51 max: 1.0

• Samples:

  query	doc	label
  tahun berapa ekspor naik 2,37% dan impor naik 30,30%?	Bulan November 2006 Ekspor Naik 2,37 % dan Impor Naik 30,30 %	1.0
  Berapa produksi padi pada tahun 2023?	Produksi padi tahun lainnya	0.0
  data statistik solus per aqua 2015	Statistik Solus Per Aqua (SPA) 2015	0.97

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• num_train_epochs: 16
• warmup_ratio: 0.1
• fp16: 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: 16
• 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: 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: False
• 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

Click to expand

Epoch	Step	Training Loss	Validation Loss	allstats-semantic-search-v1-3-dev_spearman_cosine	allstat-semantic-search-v1-3-test_spearman_cosine
0.1502	500	0.0579	0.0351	0.7132	-
0.3005	1000	0.03	0.0225	0.7589	-
0.4507	1500	0.0219	0.0185	0.7834	-
0.6010	2000	0.0181	0.0163	0.7946	-
0.7512	2500	0.0162	0.0147	0.7941	-
0.9014	3000	0.015	0.0147	0.8050	-
1.0517	3500	0.014	0.0131	0.7946	-
1.2019	4000	0.0119	0.0126	0.8038	-
1.3522	4500	0.0121	0.0128	0.8213	-
1.5024	5000	0.0117	0.0116	0.8268	-
1.6526	5500	0.0124	0.0117	0.8269	-
1.8029	6000	0.0111	0.0109	0.8421	-
1.9531	6500	0.0105	0.0108	0.8278	-
2.1034	7000	0.0091	0.0093	0.8460	-
2.2536	7500	0.0085	0.0091	0.8469	-
2.4038	8000	0.0079	0.0083	0.8595	-
2.5541	8500	0.0075	0.0085	0.8495	-
2.7043	9000	0.0073	0.0082	0.8614	-
2.8546	9500	0.0068	0.0077	0.8696	-
3.0048	10000	0.0066	0.0076	0.8669	-
3.1550	10500	0.0058	0.0072	0.8678	-
3.3053	11000	0.0056	0.0067	0.8703	-
3.4555	11500	0.0054	0.0067	0.8766	-
3.6058	12000	0.0054	0.0063	0.8678	-
3.7560	12500	0.0051	0.0061	0.8786	-
3.9062	13000	0.0052	0.0077	0.8699	-
4.0565	13500	0.005	0.0055	0.8859	-
4.2067	14000	0.0041	0.0054	0.8900	-
4.3570	14500	0.0038	0.0052	0.8892	-
4.5072	15000	0.0039	0.0050	0.8895	-
4.6575	15500	0.004	0.0052	0.8972	-
4.8077	16000	0.0042	0.0051	0.8927	-
4.9579	16500	0.0041	0.0052	0.8930	-
5.1082	17000	0.0034	0.0053	0.8998	-
5.2584	17500	0.003	0.0047	0.9023	-
5.4087	18000	0.0032	0.0045	0.9039	-
5.5589	18500	0.0032	0.0044	0.8996	-
5.7091	19000	0.0032	0.0041	0.9085	-
5.8594	19500	0.0032	0.0047	0.9072	-
6.0096	20000	0.0029	0.0037	0.9104	-
6.1599	20500	0.0024	0.0037	0.9112	-
6.3101	21000	0.0026	0.0039	0.9112	-
6.4603	21500	0.0024	0.0037	0.9157	-
6.6106	22000	0.0022	0.0038	0.9122	-
6.7608	22500	0.0025	0.0034	0.9170	-
6.9111	23000	0.0023	0.0034	0.9179	-
7.0613	23500	0.002	0.0031	0.9244	-
7.2115	24000	0.0019	0.0030	0.9250	-
7.3618	24500	0.0018	0.0032	0.9249	-
7.5120	25000	0.0022	0.0031	0.9162	-
7.6623	25500	0.0019	0.0030	0.9266	-
7.8125	26000	0.0019	0.0028	0.9297	-
7.9627	26500	0.0018	0.0028	0.9282	-
8.1130	27000	0.0015	0.0025	0.9324	-
8.2632	27500	0.0014	0.0027	0.9337	-
8.4135	28000	0.0015	0.0027	0.9327	-
8.5637	28500	0.0016	0.0027	0.9313	-
8.7139	29000	0.0016	0.0027	0.9333	-
8.8642	29500	0.0015	0.0025	0.9382	-
9.0144	30000	0.0014	0.0025	0.9375	-
9.1647	30500	0.0011	0.0024	0.9398	-
9.3149	31000	0.0012	0.0025	0.9384	-
9.4651	31500	0.0014	0.0025	0.9383	-
9.6154	32000	0.0013	0.0023	0.9410	-
9.7656	32500	0.0011	0.0023	0.9409	-
9.9159	33000	0.0012	0.0021	0.9432	-
10.0661	33500	0.0011	0.0021	0.9432	-
10.2163	34000	0.001	0.0021	0.9442	-
10.3666	34500	0.0009	0.0022	0.9436	-
10.5168	35000	0.001	0.0021	0.9468	-
10.6671	35500	0.001	0.0020	0.9471	-
10.8173	36000	0.001	0.0021	0.9467	-
10.9675	36500	0.0011	0.0021	0.9478	-
11.1178	37000	0.0008	0.0020	0.9493	-
11.2680	37500	0.0008	0.0019	0.9509	-
11.4183	38000	0.0008	0.0019	0.9504	-
11.5685	38500	0.0008	0.0019	0.9512	-
11.7188	39000	0.0008	0.0019	0.9516	-
11.8690	39500	0.0007	0.0019	0.9534	-
12.0192	40000	0.0007	0.0018	0.9539	-
12.1695	40500	0.0006	0.0018	0.9555	-
12.3197	41000	0.0006	0.0019	0.9551	-
12.4700	41500	0.0007	0.0019	0.9550	-
12.6202	42000	0.0008	0.0018	0.9552	-
12.7704	42500	0.0006	0.0017	0.9559	-
12.9207	43000	0.0006	0.0017	0.9568	-
13.0709	43500	0.0006	0.0017	0.9577	-
13.2212	44000	0.0005	0.0017	0.9581	-
13.3714	44500	0.0006	0.0017	0.9586	-
13.5216	45000	0.0005	0.0017	0.9587	-
13.6719	45500	0.0005	0.0017	0.9591	-
13.8221	46000	0.0006	0.0016	0.9600	-
13.9724	46500	0.0005	0.0016	0.9603	-
14.1226	47000	0.0005	0.0016	0.9609	-
14.2728	47500	0.0005	0.0016	0.9612	-
14.4231	48000	0.0005	0.0016	0.9611	-
14.5733	48500	0.0005	0.0016	0.9616	-
14.7236	49000	0.0004	0.0015	0.9625	-
14.8738	49500	0.0004	0.0016	0.9628	-
15.0240	50000	0.0004	0.0016	0.9631	-
15.1743	50500	0.0004	0.0016	0.9632	-
15.3245	51000	0.0004	0.0016	0.9633	-
15.4748	51500	0.0004	0.0016	0.9635	-
15.625	52000	0.0004	0.0015	0.9638	-
15.7752	52500	0.0004	0.0015	0.9640	-
15.9255	53000	0.0004	0.0015	0.9641	-
16.0	53248	-	-	-	0.9648

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.3.1
• Transformers: 4.47.1
• PyTorch: 2.2.2+cu121
• Accelerate: 1.2.1
• 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",
}