This is a sentence-transformers model finetuned from thenlper/gte-base. 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: thenlper/gte-base
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 tokens
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': 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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'And then comes the figure of the human in the age of the Anthropocene, the era when humans act as a geological force on the planet, changing its climate for millennia to come.',
    'â\x80\x98Anthropoceneâ\x80\x99 means, after all, â\x80\x98new Man time.â\x80\x99 For, while the Anthropocene, as a name, claims a generalised human agency responsible for the myriad ecological crises gathered under its auspice, it is simply not the case that, as Ghosh argues, â\x80\x9cevery human being, past and present, has contributed to the present cycle of climate changeâ\x80\x9d (2016, 115).',
    'Minneapolis: University of Minnesota Press, 2007.',
]
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]

Training Details

Training Dataset

Unnamed Dataset

Size: 134,934 training samples
Columns: inp1, inp2, and score
Approximate statistics based on the first 1000 samples:
inp1 inp2 score
type string string float
details
min: 8 tokens
mean: 38.09 tokens
max: 512 tokens
min: 8 tokens
mean: 32.43 tokens
max: 512 tokens
min: -1.0
mean: -0.8
max: 1.0

Samples:

inp1	inp2	score
`Following the lead of John Guillory in Cultural Capital: The Problem of Literary Canon Formation, I would argue that such theoretical arguments characteristically concern an âimaginary canonââimaginary in that there is no speciï¬cally deï¬ned body of works or authors that make up such a canon.`	`âBrooksâs theory,â guillory writes in Cultural Capital: The Problem of Liter- ary Canon Formation (Chicago: Univ.`	`1.0`
`Cultural Capital: The Problem of Literary Canon Formation.`	`âBrooksâs theory,â guillory writes in Cultural Capital: The Problem of Liter- ary Canon Formation (Chicago: Univ.`	`1.0`
`A partic- ularly good example of the complex operations of critical attention and peda- gogical appropriation occurs with Zora Neale Hurstonâs Their Eyes Were Watching God.`	`Similarly, in her article comparing the image patterns in Zora Neale Hurstonâs Their Eyes Were Watching God and Beloved, Glenda B. Weathers also observes the dichotomous function of the trees in Beloved and argues, âThey posit knowledge of both good and evilâ (2005, 201) for black Americans seek- ing freedom from slavery and oppression.`	`1.0`

Loss: CoSENTLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "pairwise_cos_sim" 4}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 2
warmup_ratio: 0.1
fp16: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
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
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: 2
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
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, '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_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss
0.0119	100	2.2069
0.0237	200	2.3883
0.0119	100	1.8358
0.0237	200	1.974
0.0356	300	1.8488
0.0474	400	1.8799
0.0593	500	2.0132
0.0711	600	1.8831
0.0830	700	1.601
0.0948	800	2.0316
0.1067	900	1.9483
0.1185	1000	1.6585
0.1304	1100	1.7986
0.1422	1200	1.4978
0.1541	1300	1.6035
0.1660	1400	1.9908
0.1778	1500	1.2896
0.1897	1600	1.97
0.2015	1700	1.9622
0.2134	1800	1.4706
0.2252	1900	1.5162
0.2371	2000	1.6988
0.2489	2100	1.6552
0.2608	2200	1.7779
0.2726	2300	1.9001
0.2845	2400	1.7802
0.2963	2500	1.6582
0.3082	2600	1.377
0.3201	2700	1.473
0.3319	2800	1.441
0.3438	2900	1.8727
0.3556	3000	1.1545
0.3675	3100	1.7319
0.3793	3200	1.9862
0.3912	3300	1.467
0.4030	3400	2.125
0.4149	3500	2.0474
0.4267	3600	1.7078
0.4386	3700	1.7791
0.4505	3800	1.6368
0.4623	3900	1.4451
0.4742	4000	1.5612
0.4860	4100	1.3163
0.4979	4200	1.5675
0.5097	4300	1.2766
0.5216	4400	1.4506
0.5334	4500	0.9601
0.5453	4600	1.4118
0.5571	4700	1.3951
0.5690	4800	1.2048
0.5808	4900	1.1108
0.5927	5000	1.5696
0.6046	5100	1.4223
0.6164	5200	1.1789
0.6283	5300	1.1573
0.6401	5400	1.4457
0.6520	5500	1.6622
0.6638	5600	1.2699
0.6757	5700	1.0191
0.6875	5800	1.2764
0.6994	5900	0.8999
0.6046	5100	1.5085
0.6164	5200	1.3738
0.6283	5300	1.0537
0.6401	5400	1.3578
0.6520	5500	1.6301
0.6638	5600	1.091
0.6757	5700	0.9261
0.6875	5800	1.1276
0.6994	5900	0.7678
0.6047	5100	1.2021
0.6166	5200	0.8787
0.6284	5300	0.6169
0.6403	5400	0.9881
0.6521	5500	1.1844
0.6640	5600	1.032
0.6758	5700	0.8486
0.6877	5800	1.4845
0.6995	5900	1.4
0.7114	6000	0.9685
0.7233	6100	0.9288
0.7351	6200	1.4682
0.7470	6300	0.6551
0.7588	6400	0.5513
0.7707	6500	0.6092
0.7825	6600	1.3235
0.7944	6700	0.4917
0.8063	6800	0.8944
0.8181	6900	0.9298
0.8300	7000	1.1134
0.8418	7100	0.8254
0.8537	7200	1.3363
0.8655	7300	0.6571
0.8774	7400	0.8209
0.8893	7500	0.6508
0.9011	7600	1.1972
0.9130	7700	1.1095
0.9248	7800	0.8772
0.9367	7900	1.0623
0.9485	8000	0.6073
0.9604	8100	0.8292
0.9723	8200	0.6765
0.9841	8300	0.5103
0.9960	8400	1.0618
1.0078	8500	0.5134
1.0197	8600	0.5203
1.0315	8700	0.6634
1.0434	8800	0.6644
1.0553	8900	0.7459
1.0671	9000	0.5969
1.0790	9100	0.5473
1.0908	9200	0.5495
1.1027	9300	0.5093
1.1145	9400	0.7049
1.1264	9500	0.726
1.1382	9600	0.6512
1.1501	9700	0.5121
1.1620	9800	0.5977
1.1738	9900	0.4933
1.1857	10000	0.8585
1.1975	10100	0.2955
1.2094	10200	0.6972
1.2212	10300	0.454
1.2331	10400	1.1057
1.2450	10500	0.9724
1.2568	10600	0.3057
1.2687	10700	0.5967
1.2805	10800	0.7332
1.2924	10900	0.5382
1.3042	11000	0.625
1.3161	11100	0.5354
1.3280	11200	0.4289
1.3398	11300	0.4243
1.3517	11400	0.6902
1.3635	11500	0.4248
1.3754	11600	0.3743
1.3872	11700	0.5463
1.3991	11800	0.8413
1.4110	11900	0.4748
1.4228	12000	0.56
1.4347	12100	0.9269
1.4465	12200	0.4668
1.4584	12300	0.4842
1.4702	12400	0.5172
1.4821	12500	0.4498
1.4940	12600	0.4695
1.5058	12700	0.2144
1.5177	12800	0.8002
1.5295	12900	0.4022
1.5414	13000	0.4491
1.5532	13100	0.4798
1.5651	13200	0.7489
1.5770	13300	0.6108
1.5888	13400	0.3806
1.6007	13500	0.4164
1.6125	13600	0.6362
1.6244	13700	0.4773
1.6362	13800	0.4875
1.6481	13900	0.5577
1.6599	14000	0.3318
1.6718	14100	0.2959
1.6837	14200	0.3168
1.6955	14300	0.403
1.7074	14400	0.6553
1.7192	14500	0.5814
1.7311	14600	0.3407
1.7429	14700	0.3985
1.7548	14800	0.406
1.7667	14900	0.5986
1.7785	15000	0.7694
1.7904	15100	0.5025
1.8022	15200	0.7199
1.8141	15300	0.4215
1.8259	15400	0.5484
1.8378	15500	0.3551
1.8497	15600	0.3572
1.8615	15700	0.3536
1.8734	15800	0.5116
1.8852	15900	0.7094
1.8971	16000	0.4402
1.9089	16100	0.4095
1.9208	16200	0.2173
1.9327	16300	0.6058
1.9445	16400	0.7796
1.9564	16500	0.5642
1.9682	16600	0.3085
1.9801	16700	0.4308
1.9919	16800	0.3712

Framework Versions

Python: 3.11.5
Sentence Transformers: 3.0.1
Transformers: 4.40.0
PyTorch: 2.2.2
Accelerate: 0.31.0
Datasets: 2.19.2
Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

bibtex

@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