SentenceTransformer based on thenlper/gte-small

This is a sentence-transformers model finetuned from thenlper/gte-small. 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: thenlper/gte-small
Maximum Sequence Length: 512 tokens
Output Dimensionality: 384 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': 512, '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})
)

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("Alexhuou/embedder_model_FT")
# Run inference
sentences = [
    'This question refers to the following information.\n"Those whose condition is such that their function is the use of their bodies and nothing better can be expected of them, those, I say, are slaves of nature. It is better for them to be ruled thus."\nJuan de Sepulveda, Politics, 1522\n"When Latin American nations gained independence in the 19th century, those two strains converged, and merged with an older, more universalist, natural law tradition. The result was a distinctively Latin American form of rights discourse. Paolo Carozza traces the roots of that discourse to a distinctive application, and extension, of Thomistic moral philosophy to the injustices of Spanish conquests in the New World. The key figure in that development seems to have been Bartolomé de Las Casas, a 16th-century Spanish bishop who condemned slavery and championed the cause of Indians on the basis of a natural right to liberty grounded in their membership in a single common humanity. \'All the peoples of the world are humans,\' Las Casas wrote, and \'all the races of humankind are one.\' According to Brian Tierney, Las Casas and other Spanish Dominican philosophers laid the groundwork for a doctrine of natural rights that was independent of religious revelation \'by drawing on a juridical tradition that derived natural rights and natural law from human rationality and free will, and by appealing to Aristotelian philosophy.\'"\nMary Ann Glendon, "The Forgotten Crucible: The Latin American Influence on the Universal Human Rights Idea,” 2003\nWhich one of the following statements about the Spanish conquest of the Americas is most accurate?',
    'This question refers to the following information.\n"One-half of the people of this nation to-day are utterly powerless to blot from the statute books an unjust law, or to write there a new and a just one. The women, dissatisfied as they are with this form of government, that enforces taxation without representation,—that compels them to obey laws to which they have never given their consent,—that imprisons and hangs them without a trial by a jury of their peers, that robs them, in marriage, of the custody of their own persons, wages and children,—are this half of the people left wholly at the mercy of the other half, in direct violation of the spirit and letter of the declarations of the framers of this government, every one of which was based on the immutable principle of equal rights to all."\n—Susan B. Anthony, "I Stand Before You Under Indictment" (speech), 1873\nWhich of the following statements best represents the criticism of Andrew Carnegie found in this cartoon?',
    'If the finite group G contains a subgroup of order seven but no element (other than the identity) is its own inverse, then the order of G could be',
]
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]

Training Details

Training Dataset

Unnamed Dataset

Size: 5,700 training samples
Columns: sentence_0, sentence_1, and sentence_2

Approximate statistics based on the first 1000 samples:

	sentence_0	sentence_1	sentence_2
type	string	string	string
details	min: 3 tokens mean: 44.65 tokens max: 512 tokens	min: 3 tokens mean: 44.72 tokens max: 512 tokens	min: 3 tokens mean: 47.14 tokens max: 512 tokens

Samples:

sentence_0	sentence_1	sentence_2
This question refers to the following information. "When the Portuguese go from Macao in China to Japan, they carry much white silk, gold, musk, and porcelain: and they bring from Japan nothing but silver. They have a great carrack which goes there every year and she brings from there every year about six hundred coins: and all this silver of Japan, and two hundred thousand coins more in silver which they bring yearly out of India, they employ to their great advantage in China: and they bring from there gold, musk, silk, copper, porcelains, and many other things very costly and gilded. When the Portuguese come to Canton in China to traffic, they must remain there but certain days: and when they come in at the gate of the city, they must enter their names in a book, and when they go out at night they must put out their names. They may not lie in the town all night, but must lie in their boats outside of the town. And, their time expired, if any man remains there, he is imprisoned." Ralp...	This question refers to the following information. Although in Protestant Europe, [Peter the Great] was surrounded by evidence of the new civil and political rights of individual men embodied in constitutions, bills of rights and parliaments, he did not return to Russia determined to share power with his people. On the contrary, he returned not only determined to change his country but also convinced that if Russia was to be transformed, it was he who must provide both the direction and the motive force. He would try to lead; but where education and persuasion were not enough, he could drive—and if necessary flog—the backward nation forward. —Robert K. Massie, Peter the Great: His Life and World Based on the above passage, what kinds of reforms did Peter the Great embrace?	This question refers to the following information. Now, we have organized a society, and we call it "Share Our Wealth Society," a society with the motto "Every Man a King."… We propose to limit the wealth of big men in the country. There is an average of $15,000 in wealth to every family in America. That is right here today. We do not propose to divide it up equally. We do not propose a division of wealth, but we do propose to limit poverty that we will allow to be inflicted on any man's family. We will not say we are going to try to guarantee any equality … but we do say that one third of the average is low enough for any one family to hold, that there should be a guarantee of a family wealth of around $5,000; enough for a home, an automobile, a radio, and the ordinary conveniences, and the opportunity to educate their children.… We will have to limit fortunes. Our present plan is that we will allow no man to own more than $50,000,000. We think that with that limit we will be able to ...
This question refers to the following information. An Act to place certain restrictions on Immigration and to provide for the removal from the Commonwealth of Prohibited Immigrants. … 3. The immigration into the Commonwealth of the persons described in any of the following paragraphs in this section (hereinafter called "prohibited immigrants") is prohibited, namely (a) Any person who when asked to do so by an officer fails to write out at dictation and sign in the presence of the officer a passage of fifty words in length in a European language directed by the officer; (b) Any person in the opinion of the Minister or of an officer to become a charge upon the public or upon any public or charitable organisation; … (g) Any persons under a contract or agreement to perform manual labour within the Commonwealth: Provided that this paragraph shall not apply to workmen exempted by the Minister for special skill required by Australia… Immigration Restriction Act of 1901 (Australia) Whereas in ...	This question refers to the following information. "My little homestead in the city, which I recently insured for ￡2,000 would no doubt have shared the common fate, as the insurance companies will not make good that which is destroyed by the Queen's enemies. And although I have a farm of 50 acres close to the town, no doubt the crops and premises would have been destroyed. In fact, this has already partly been the case, and I am now suing the Government for damages done by a contingent of 1,500 natives that have recently encamped not many hundred yards from the place, who have done much damage all around." Letter from a British citizen to his sister during the Anglo-Zulu War, South Africa, 1879 Incidents such as those described by the author of the letter were used by the British government to do which of the following?	`If the price of a product decreases with the price of a substitute product remaining constant such that the consumer buys more of this product, this is called the`
`The term 'marketing mix' describes:`	`____________________ are those who address the same target market but provide a different offering to satisfy the market need, for example Spotify, Sony, and Apple's iPod.`	`The anatomic location of the spinal canal is`

Loss: TripletLoss with these parameters:

{
    "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
    "triplet_margin": 5
}

Training Hyperparameters

Non-Default Hyperparameters

multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 8
per_device_eval_batch_size: 8
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
num_train_epochs: 3
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
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: 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: 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
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: round_robin

Training Logs

Epoch	Step	Training Loss
1.4006	500	1.7059
2.8011	1000	0.88
0.7013	500	0.7526
1.4025	1000	0.6176
2.1038	1500	0.4602
2.8050	2000	0.3472

Framework Versions

Python: 3.11.13
Sentence Transformers: 4.1.0
Transformers: 4.52.4
PyTorch: 2.6.0+cu124
Accelerate: 1.7.0
Datasets: 3.6.0
Tokenizers: 0.21.1

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",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification},
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}

Alexhuou
/

embedder_model_FT