metadata

language:
  - id
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:6198
  - loss:CoSENTLoss
base_model: intfloat/multilingual-e5-base
datasets:
  - Pustekhan-ITB/stsb-indo-edu
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - pearson_cosine
  - spearman_cosine
model-index:
  - name: SentenceTransformer based on intfloat/multilingual-e5-base
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: stsb indo edu dev
          type: stsb-indo-edu-dev
        metrics:
          - type: pearson_cosine
            value: 0.1930033858243812
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.17647076252403324
            name: Spearman Cosine
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: stsb indo edu test
          type: stsb-indo-edu-test
        metrics:
          - type: pearson_cosine
            value: 0.15065000397563194
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.1512326380689479
            name: Spearman Cosine

SentenceTransformer based on intfloat/multilingual-e5-base

This is a sentence-transformers model finetuned from intfloat/multilingual-e5-base on the stsb-indo-edu 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: intfloat/multilingual-e5-base
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- stsb-indo-edu
Language: id

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}) 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})
  (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("ewideplus/indoedu-e5-base")
# Run inference
sentences = [
    'The weather is lovely today.',
    "It's so sunny outside!",
    'He drove to the stadium.',
]
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: stsb-indo-edu-dev and stsb-indo-edu-test
Evaluated with EmbeddingSimilarityEvaluator

Metric	stsb-indo-edu-dev	stsb-indo-edu-test
pearson_cosine	0.193	0.1507
spearman_cosine	0.1765	0.1512

Training Details

Training Dataset

stsb-indo-edu

Dataset: stsb-indo-edu at f84d4d6
Size: 6,198 training samples
Columns: sentence1, sentence2, and score

Approximate statistics based on the first 1000 samples:

	sentence1	sentence2	score
type	list	list	float
details	min: 18 elements mean: 58.40 elements max: 137 elements	min: 15 elements mean: 54.31 elements max: 118 elements	min: 0.0 mean: 0.46 max: 1.0

Samples:

sentence1	sentence2	score
`['query: P', 'query: e', 'query: l', 'query: a', 'query: j', ...]`	`['passage: T', 'passage: a', 'passage: r', 'passage: i', 'passage: a', ...]`	`0.76`
`['query: S', 'query: e', 'query: b', 'query: e', 'query: l', ...]`	`['passage: U', 'passage: p', 'passage: a', 'passage: y', 'passage: a', ...]`	`0.85`
`['query: B', 'query: e', 'query: b', 'query: e', 'query: r', ...]`	`['passage: I', 'passage: n', 'passage: i', 'passage: ', 'passage: m', ...]`	`0.63`

Loss: CoSENTLoss with these parameters:

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

Evaluation Dataset

stsb-indo-edu

Dataset: stsb-indo-edu at f84d4d6
Size: 1,536 evaluation samples
Columns: sentence1, sentence2, and score

Approximate statistics based on the first 1000 samples:

	sentence1	sentence2	score
type	list	list	float
details	min: 14 elements mean: 86.67 elements max: 172 elements	min: 22 elements mean: 88.94 elements max: 177 elements	min: 0.0 mean: 0.42 max: 1.0

Samples:

sentence1	sentence2	score
`['query: S', 'query: e', 'query: o', 'query: r', 'query: a', ...]`	`['passage: S', 'passage: e', 'passage: o', 'passage: r', 'passage: a', ...]`	`1.0`
`['query: S', 'query: e', 'query: o', 'query: r', 'query: a', ...]`	`['passage: S', 'passage: e', 'passage: o', 'passage: r', 'passage: a', ...]`	`0.95`
`['query: S', 'query: e', 'query: o', 'query: r', 'query: a', ...]`	`['passage: P', 'passage: r', 'passage: i', 'passage: a', 'passage: ', ...]`	`1.0`

Loss: CoSENTLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 32
per_device_eval_batch_size: 16
learning_rate: 1e-05
weight_decay: 0.01
num_train_epochs: 5
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: 32
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: 1e-05
weight_decay: 0.01
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: 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

Epoch	Step	Training Loss	Validation Loss	stsb-indo-edu-dev_spearman_cosine	stsb-indo-edu-test_spearman_cosine
-1	-1	-	-	0.0995	-
0.5155	100	6.2244	4.7594	0.1027	-
1.0309	200	6.1605	4.7518	0.1502	-
1.5464	300	6.16	4.7553	0.1564	-
2.0619	400	6.1609	4.7527	0.1714	-
2.5773	500	6.1593	4.7698	0.1495	-
3.0928	600	6.1517	4.7516	0.1657	-
3.6082	700	6.1555	4.7463	0.1787	-
4.1237	800	6.1452	4.7548	0.1665	-
4.6392	900	6.1523	4.7494	0.1765	-
-1	-1	-	-	-	0.1512

Framework Versions

Python: 3.11.11
Sentence Transformers: 3.4.1
Transformers: 4.48.3
PyTorch: 2.5.1+cu124
Accelerate: 1.3.0
Datasets: 3.3.2
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",
}

CoSENTLoss

@online{kexuefm-8847,
    title={CoSENT: A more efficient sentence vector scheme than Sentence-BERT},
    author={Su Jianlin},
    year={2022},
    month={Jan},
    url={https://kexue.fm/archives/8847},
}