SentenceTransformer based on sentence-transformers/all-mpnet-base-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-mpnet-base-v2. 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/all-mpnet-base-v2
Maximum Sequence Length: 384 tokens
Output Dimensionality: 768 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': 384, 'do_lower_case': False}) with Transformer model: MPNetModel 
  (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("sentence_transformers_model_id")
# Run inference
sentences = [
    "Background: Patients find it difficult to use dexterous prosthetic hands without a suitable control system, highlighting a need for improved grasp performance and ease of operation. Existing methods may not adequately address the challenges faced by users, particularly those with inferior myoelectric signals, in effectively controlling prosthetic devices.\nContribution: Combine 'myoelectric signal' and ",
    'a unified framework for collaborative decoding between large and small language models (Large Language Models and small language models)',
    'joint biomedical entity linking and event extraction',
]
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: 784,827 training samples
Columns: query, answer, and label
Approximate statistics based on the first 1000 samples:
query answer label
type string string int
details
min: 60 tokens
mean: 77.86 tokens
max: 93 tokens

min: 3 tokens
mean: 8.82 tokens
max: 64 tokens

0: ~96.70%
1: ~3.30%

	query	answer	label
type	string	string	int
details	min: 60 tokens mean: 77.86 tokens max: 93 tokens	min: 3 tokens mean: 8.82 tokens max: 64 tokens	0: ~96.70% 1: ~3.30%

Samples:

query	answer	label
`Background: The study addresses the challenge of action segmentation under weak supervision, where the available ground truth only indicates the presence of actions without providing their temporal ordering or occurrence timing in training videos. This limitation necessitates the development of a method to generate pseudo-ground truth for effective training and improve performance in action segmentation and alignment tasks. Contribution: Combine 'a Hidden Markov Model' and`	`a multilayer perceptron`	`1`
`Background: The study addresses the challenge of action segmentation under weak supervision, where the available ground truth only indicates the presence of actions without providing their temporal ordering or occurrence timing in training videos. This limitation necessitates the development of a method to generate pseudo-ground truth for effective training and improve performance in action segmentation and alignment tasks. Contribution: Combine 'a Hidden Markov Model' and`	`synthetic occlusion augmentation during training`	`0`
`Background: The study addresses the challenge of action segmentation under weak supervision, where the available ground truth only indicates the presence of actions without providing their temporal ordering or occurrence timing in training videos. This limitation necessitates the development of a method to generate pseudo-ground truth for effective training and improve performance in action segmentation and alignment tasks. Contribution: Combine 'a Hidden Markov Model' and`	`robustness of deep learning methods`	`0`

Loss: ContrastiveLoss with these parameters:

{
    "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
    "margin": 0.5,
    "size_average": true
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 64
learning_rate: 1.9218937402834593e-05
num_train_epochs: 2
warmup_ratio: 0.08278167292320517
bf16: True
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 64
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: 1.9218937402834593e-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: 0.08278167292320517
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: True
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
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: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss
0.0082	100	0.0104
0.0163	200	0.0068
0.0245	300	0.005
0.0326	400	0.0041
0.0408	500	0.0054
0.0489	600	0.004
0.0571	700	0.0037
0.0652	800	0.0037
0.0734	900	0.0049
0.0815	1000	0.0038
0.0897	1100	0.004
0.0979	1200	0.0037
0.1060	1300	0.004
0.1142	1400	0.0049
0.1223	1500	0.0038
0.1305	1600	0.0036
0.1386	1700	0.0037
0.1468	1800	0.0045
0.1549	1900	0.0038
0.1631	2000	0.0034
0.1712	2100	0.0034
0.1794	2200	0.0035
0.1876	2300	0.0045
0.1957	2400	0.0036
0.2039	2500	0.0036
0.2120	2600	0.0033
0.2202	2700	0.004
0.2283	2800	0.0036
0.2365	2900	0.0033
0.2446	3000	0.0033
0.2528	3100	0.0037
0.2609	3200	0.0038
0.2691	3300	0.0033
0.2773	3400	0.0034
0.2854	3500	0.0033
0.2936	3600	0.0041
0.3017	3700	0.0033
0.3099	3800	0.0033
0.3180	3900	0.0032
0.3262	4000	0.004
0.3343	4100	0.0035
0.3425	4200	0.0031
0.3506	4300	0.0033
0.3588	4400	0.0033
0.3670	4500	0.0039
0.3751	4600	0.0032
0.3833	4700	0.0034
0.3914	4800	0.0031
0.3996	4900	0.004
0.4077	5000	0.0032
0.4159	5100	0.0031
0.4240	5200	0.0031
0.4322	5300	0.0032
0.4403	5400	0.0039
0.4485	5500	0.0031
0.4567	5600	0.003
0.4648	5700	0.0032
0.4730	5800	0.0038
0.4811	5900	0.0033
0.4893	6000	0.0031
0.4974	6100	0.0032
0.5056	6200	0.0033
0.5137	6300	0.0033
0.5219	6400	0.0032
0.5300	6500	0.0031
0.5382	6600	0.0032
0.5464	6700	0.0038
0.5545	6800	0.003
0.5627	6900	0.003
0.5708	7000	0.0029
0.5790	7100	0.0038
0.5871	7200	0.0032
0.5953	7300	0.0031
0.6034	7400	0.003
0.6116	7500	0.003
0.6198	7600	0.0039
0.6279	7700	0.0031
0.6361	7800	0.0031
0.6442	7900	0.0031
0.6524	8000	0.0039
0.6605	8100	0.003
0.6687	8200	0.003
0.6768	8300	0.003
0.6850	8400	0.0028
0.6931	8500	0.0035
0.7013	8600	0.0031
0.7095	8700	0.003
0.7176	8800	0.0026
0.7258	8900	0.0034
0.7339	9000	0.0033
0.7421	9100	0.003
0.7502	9200	0.0027
0.7584	9300	0.0029
0.7665	9400	0.0034
0.7747	9500	0.0029
0.7828	9600	0.0028
0.7910	9700	0.0027
0.7992	9800	0.0033
0.8073	9900	0.0031
0.8155	10000	0.0029
0.8236	10100	0.0028
0.8318	10200	0.0031
0.8399	10300	0.0031
0.8481	10400	0.003
0.8562	10500	0.0029
0.8644	10600	0.0028
0.8725	10700	0.0033
0.8807	10800	0.003
0.8889	10900	0.0029
0.8970	11000	0.0027
0.9052	11100	0.0033
0.9133	11200	0.0029
0.9215	11300	0.0029
0.9296	11400	0.0029
0.9378	11500	0.003
0.9459	11600	0.0034
0.9541	11700	0.0031
0.9622	11800	0.0027
0.9704	11900	0.0029
0.9786	12000	0.0034
0.9867	12100	0.0032
0.9949	12200	0.003
1.0030	12300	0.0032
1.0112	12400	0.0028
1.0193	12500	0.003
1.0275	12600	0.0027
1.0356	12700	0.0034
1.0438	12800	0.0029
1.0519	12900	0.0025
1.0601	13000	0.0028
1.0683	13100	0.0026
1.0764	13200	0.0035
1.0846	13300	0.0026
1.0927	13400	0.0028
1.1009	13500	0.0026
1.1090	13600	0.0034
1.1172	13700	0.0028
1.1253	13800	0.0027
1.1335	13900	0.0026
1.1416	14000	0.0031
1.1498	14100	0.0025
1.1580	14200	0.0025
1.1661	14300	0.0025
1.1743	14400	0.0024
1.1824	14500	0.0031
1.1906	14600	0.0025
1.1987	14700	0.0024
1.2069	14800	0.0025
1.2150	14900	0.0029
1.2232	15000	0.0025
1.2313	15100	0.0025
1.2395	15200	0.0023
1.2477	15300	0.0024
1.2558	15400	0.0029
1.2640	15500	0.0023
1.2721	15600	0.0023
1.2803	15700	0.0023
1.2884	15800	0.0032
1.2966	15900	0.0023
1.3047	16000	0.0023
1.3129	16100	0.0024
1.3210	16200	0.0025
1.3292	16300	0.0028
1.3374	16400	0.0023
1.3455	16500	0.0021
1.3537	16600	0.0023
1.3618	16700	0.0029
1.3700	16800	0.0023
1.3781	16900	0.0023
1.3863	17000	0.0025
1.3944	17100	0.0028
1.4026	17200	0.0023
1.4107	17300	0.0023
1.4189	17400	0.0023
1.4271	17500	0.0023
1.4352	17600	0.0029
1.4434	17700	0.0022
1.4515	17800	0.0022
1.4597	17900	0.0023
1.4678	18000	0.0026
1.4760	18100	0.0024
1.4841	18200	0.0023
1.4923	18300	0.0024
1.5004	18400	0.0024
1.5086	18500	0.0026
1.5168	18600	0.0022
1.5249	18700	0.0023
1.5331	18800	0.0023
1.5412	18900	0.003
1.5494	19000	0.002
1.5575	19100	0.0022
1.5657	19200	0.0023
1.5738	19300	0.0023
1.5820	19400	0.0028
1.5901	19500	0.0022
1.5983	19600	0.0023
1.6065	19700	0.0022
1.6146	19800	0.0028
1.6228	19900	0.0022
1.6309	20000	0.0023
1.6391	20100	0.0025
1.6472	20200	0.0028
1.6554	20300	0.0023
1.6635	20400	0.0021
1.6717	20500	0.0022
1.6798	20600	0.0022
1.6880	20700	0.0025
1.6962	20800	0.0024
1.7043	20900	0.0023
1.7125	21000	0.0021
1.7206	21100	0.0024
1.7288	21200	0.0024
1.7369	21300	0.0023
1.7451	21400	0.0022
1.7532	21500	0.0021
1.7614	21600	0.0025
1.7696	21700	0.0023
1.7777	21800	0.002
1.7859	21900	0.0022
1.7940	22000	0.0025
1.8022	22100	0.0022
1.8103	22200	0.0023
1.8185	22300	0.0022
1.8266	22400	0.0021
1.8348	22500	0.0025
1.8429	22600	0.0025
1.8511	22700	0.0022
1.8593	22800	0.0023
1.8674	22900	0.0026
1.8756	23000	0.0022
1.8837	23100	0.0022
1.8919	23200	0.0022
1.9000	23300	0.0024
1.9082	23400	0.0022
1.9163	23500	0.0022
1.9245	23600	0.0023
1.9326	23700	0.0023
1.9408	23800	0.0027
1.9490	23900	0.0023
1.9571	24000	0.0023
1.9653	24100	0.0022
1.9734	24200	0.0027
1.9816	24300	0.0025
1.9897	24400	0.0023
1.9979	24500	0.0025

Framework Versions

Python: 3.11.2
Sentence Transformers: 3.3.1
Transformers: 4.49.0
PyTorch: 2.5.1+cu124
Accelerate: 1.0.1
Datasets: 3.1.0
Tokenizers: 0.21.0

Citation

BibTeX

@misc{sternlicht2025chimeraknowledgebaseidea,
      title={CHIMERA: A Knowledge Base of Idea Recombination in Scientific Literature}, 
      author={Noy Sternlicht and Tom Hope},
      year={2025},
      eprint={2505.20779},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2505.20779}, 
}

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",
}

ContrastiveLoss

@inproceedings{hadsell2006dimensionality,
    author={Hadsell, R. and Chopra, S. and LeCun, Y.},
    booktitle={2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)},
    title={Dimensionality Reduction by Learning an Invariant Mapping},
    year={2006},
    volume={2},
    number={},
    pages={1735-1742},
    doi={10.1109/CVPR.2006.100}
}

Quick Links

🌐 Project
📃 Paper
🛠️ Code

noystl
/

recomb-pred-all-mpnet-base