SentenceTransformer based on google-bert/bert-base-uncased

This is a sentence-transformers model finetuned from google-bert/bert-base-uncased. 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: google-bert/bert-base-uncased
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})
)

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("jameswright/ws-wr-questions-bert-TSDAE-v1")
# Run inference
sentences = [
    'You have been sick so If questions it just - ’ being sorted, thanks Then assertiveness course and a . overtime isn selfish it just doing ’ s right for Why you?',
    'You have been off sick so it makes sense. If anyone questions it just grey rock - it’s being sorted, thanks. Then do an online assertiveness course and ask your GP for a CBT referral. Not doing overtime isn’t selfish - it’s just you doing what’s right for you. Why would you do anything else?',
    'Science works by the accumulation of evidence.\xa0 Independent groups work on projects and publish results.\xa0 Those results are examined and tested and examined again and tested again, such that they\'re either confirmed or discarded and further work continues accordingly.\xa0 If a scientist or doctor disagrees with the \'official line\' they\'re asked to present the data, methods and conclusions that have led to that disagreement so that it can by examined by the broader scientific and medical community.\xa0 \xa0 And yes, someone who goes on YouTube or wherever - whether doctor, scientist or layperson - and tells viewers that a vaccine alters DNA structure and destroys the immune system is either a grifter or a fruitcake. 1 minute ago, FIRETHORN1 said: ...Can you not accept that some people can hold an opposite view quite genuinely? To me, a "conspiracy theorist" is someone who believes what they are told, without any evidence to back it up. I can absolutely accept that someone can genuinely believe something without having any evidence at all to support that view.\xa0 I\'m believing that right now, in fact. 1 minute ago, FIRETHORN1 said: There is no evidence whatsoever that the vaccine works That is categorically, absolutely and undeniably false, as the most cursory of research will tell you.\xa0 But then you don\'t actually want\xa0 to believe me, do you?',
]
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: 132,712 training samples
Columns: sentence_0 and sentence_1
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1
type string string
details
min: 4 tokens
mean: 47.75 tokens
max: 460 tokens

min: 17 tokens
mean: 114.94 tokens
max: 512 tokens

	sentence_0	sentence_1
type	string	string
details	min: 4 tokens mean: 47.75 tokens max: 460 tokens	min: 17 tokens mean: 114.94 tokens max: 512 tokens

Samples:

sentence_0	sentence_1
`’ Can really go to the doctors I ’ bored of the ” . Feels more like than a doctor, does sound depression, so seeing GP a first`	`PetersRabbitt I don’t know? Can I really go to the doctors and say “hey, yes my problem is I’m bored all of the time”. Feels more like a me problem than one a doctor can help with. Yes, absolutely. It does sound like it could be depression, so seeing your GP is a good first step.`
`Ursuladevine Between 11 16, if hasn t, what has been been providing education have LakieLady Yesterday 15:34 My that dismissed offhand son the school have up for assessment . Within years referred diagnosed with PTSD,,, social anxiety and, decided his be.`	Ursuladevine · Yesterday 15:42 Between 11 and 16, if he hasn’t been attending school, what has he been doing? Has the LA been providing any education or have you been HE? LakieLady · Yesterday 15:34 My friend tried that, and the GP dismissed it offhand, saying that if her son was neurodivergent, the school would have picked up on it and referred him for assessment. Her DS was eight at the time. Within the next 2-3 years, he got much worse, was referred to CAMHS, diagnosed with significant MH problems (PTSD, GAD, depression, social anxiety disorder) and after a couple of years, CAMHS decided his mother might not be talking bollocks and that he might have ASD.
`It sounds you were a child then came along realised here was he - and this to it . young, I'd imagine`	`It sounds like you were hurt by one man when you were a child, then another came along and realised here was someone damaged he could dominate - and added his own abuse. They can sniff this out and are attracted to it. How old were you when he arrived? Very young, I'd imagine. Stepfather?`

Loss: DenoisingAutoEncoderLoss

Training Hyperparameters

Non-Default Hyperparameters

num_train_epochs: 5
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
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: 5
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: 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_eval_metrics: False
eval_on_start: False
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin

Training Logs

Click to expand

Epoch	Step	Training Loss
0.0301	500	4.7687
0.0603	1000	4.2523
0.0904	1500	4.1156
0.1206	2000	4.0278
0.1507	2500	3.9652
0.1808	3000	3.919
0.2110	3500	3.8629
0.2411	4000	3.7985
0.2713	4500	3.7625
0.3014	5000	3.7523
0.3315	5500	3.7316
0.3617	6000	3.6837
0.3918	6500	3.669
0.4220	7000	3.6394
0.4521	7500	3.6017
0.4822	8000	3.5693
0.5124	8500	3.5821
0.5425	9000	3.5488
0.5727	9500	3.5139
0.6028	10000	3.5119
0.6329	10500	3.4988
0.6631	11000	3.4741
0.6932	11500	3.4719
0.7234	12000	3.4501
0.7535	12500	3.4353
0.7837	13000	3.4107
0.8138	13500	3.4023
0.8439	14000	3.3902
0.8741	14500	3.3697
0.9042	15000	3.3731
0.9344	15500	3.3603
0.9645	16000	3.3284
0.9946	16500	3.3339
1.0248	17000	3.2793
1.0549	17500	3.2098
1.0851	18000	3.1994
1.1152	18500	3.1801
1.1453	19000	3.1634
1.1755	19500	3.1566
1.2056	20000	3.1205
1.2358	20500	3.1064
1.2659	21000	3.1028
1.2960	21500	3.099
1.3262	22000	3.1028
1.3563	22500	3.0653
1.3865	23000	3.044
1.4166	23500	3.0481
1.4467	24000	3.0133
1.4769	24500	2.9667
1.5070	25000	3.0226
1.5372	25500	2.991
1.5673	26000	2.9593
1.5974	26500	2.9598
1.6276	27000	2.9572
1.6577	27500	2.9579
1.6879	28000	2.9303
1.7180	28500	2.948
1.7481	29000	2.918
1.7783	29500	2.9014
1.8084	30000	2.8948
1.8386	30500	2.8916
1.8687	31000	2.8787
1.8988	31500	2.8864
1.9290	32000	2.8649
1.9591	32500	2.8419
1.9893	33000	2.8688
2.0194	33500	2.8329
2.0496	34000	2.7442
2.0797	34500	2.7501
2.1098	35000	2.7466
2.1400	35500	2.7343
2.1701	36000	2.7014
2.2003	36500	2.6891
2.2304	37000	2.6819
2.2605	37500	2.6779
2.2907	38000	2.6872
2.3208	38500	2.6758
2.3510	39000	2.6665
2.3811	39500	2.6392
2.4112	40000	2.6362
2.4414	40500	2.6038
2.4715	41000	2.5535
2.5017	41500	2.6081
2.5318	42000	2.6071
2.5619	42500	2.5571
2.5921	43000	2.5774
2.6222	43500	2.5556
2.6524	44000	2.5683
2.6825	44500	2.5317
2.7126	45000	2.5509
2.7428	45500	2.5292
2.7729	46000	2.52
2.8031	46500	2.4818
2.8332	47000	2.5258
2.8633	47500	2.482
2.8935	48000	2.5038
2.9236	48500	2.4864
2.9538	49000	2.4591
2.9839	49500	2.4887
3.0140	50000	2.4635
3.0442	50500	2.3837
3.0743	51000	2.3886
3.1045	51500	2.3836
3.1346	52000	2.38
3.1647	52500	2.3456
3.1949	53000	2.3171
3.2250	53500	2.3341
3.2552	54000	2.3228
3.2853	54500	2.3459
3.3154	55000	2.3251
3.3456	55500	2.3365
3.3757	56000	2.2838
3.4059	56500	2.3042
3.4360	57000	2.2465
3.4662	57500	2.2304
3.4963	58000	2.251
3.5264	58500	2.2727
3.5566	59000	2.2324
3.5867	59500	2.2325
3.6169	60000	2.2246
3.6470	60500	2.2287
3.6771	61000	2.2067
3.7073	61500	2.2206
3.7374	62000	2.1882
3.7676	62500	2.1889
3.7977	63000	2.1559
3.8278	63500	2.2021
3.8580	64000	2.1643
3.8881	64500	2.145
3.9183	65000	2.1707
3.9484	65500	2.1349
3.9785	66000	2.1659
4.0087	66500	2.152
4.0388	67000	2.0801
4.0690	67500	2.0729
4.0991	68000	2.0676
4.1292	68500	2.0622
4.1594	69000	2.0376
4.1895	69500	2.027
4.2197	70000	2.0227
4.2498	70500	2.0146
4.2799	71000	2.0334
4.3101	71500	2.0428
4.3402	72000	2.034
4.3704	72500	1.9907
4.4005	73000	2.0106
4.4306	73500	1.9488
4.4608	74000	1.961
4.4909	74500	1.9351
4.5211	75000	1.9875
4.5512	75500	1.9454
4.5813	76000	1.9453
4.6115	76500	1.9239
4.6416	77000	1.9664
4.6718	77500	1.906
4.7019	78000	1.9256
4.7321	78500	1.9071
4.7622	79000	1.9117
4.7923	79500	1.8817
4.8225	80000	1.9101
4.8526	80500	1.8872
4.8828	81000	1.8634
4.9129	81500	1.8791
4.9430	82000	1.8801
4.9732	82500	1.8586

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.0.1
Transformers: 4.42.3
PyTorch: 2.3.0+cu121
Accelerate: 0.31.0
Datasets: 2.20.0
Tokenizers: 0.19.1

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",
}

DenoisingAutoEncoderLoss

@inproceedings{wang-2021-TSDAE,
    title = "TSDAE: Using Transformer-based Sequential Denoising Auto-Encoderfor Unsupervised Sentence Embedding Learning",
    author = "Wang, Kexin and Reimers, Nils and Gurevych, Iryna", 
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2021",
    month = nov,
    year = "2021",
    address = "Punta Cana, Dominican Republic",
    publisher = "Association for Computational Linguistics",
    pages = "671--688",
    url = "https://arxiv.org/abs/2104.06979",
}