SentenceTransformer based on answerdotai/ModernBERT-base

This is a sentence-transformers model finetuned from answerdotai/ModernBERT-base on the tekgen-ctsp 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: answerdotai/ModernBERT-base
Maximum Sequence Length: 8192 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- tekgen-ctsp

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': 8192, 'do_lower_case': False}) with Transformer model: ModernBertModel 
  (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("YesaOuO/ModernBERT-base-CTSP")
# Run inference
sentences = [
    'In 1842 Alvars married the harpist Melanie Lewy, a member of a Vienna-based family of musicians with whom Alvars frequently performed.',
    'Elias Parish Alvars spouse Melanie Lewy, place of death Vienna.',
    'Elias Parish Alvars place of birth Teignmouth.',
]
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

Triplet

Dataset: YesaOuO/TEKGEN-CTSP
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9166

Training Details

Training Dataset

tekgen-ctsp

Dataset: tekgen-ctsp at 8d091eb
Size: 1,136,292 training samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative
type	string	string	string
details	min: 11 tokens mean: 38.01 tokens max: 128 tokens	min: 7 tokens mean: 17.07 tokens max: 47 tokens	min: 7 tokens mean: 17.02 tokens max: 47 tokens

Samples:

anchor	positive	negative
`1976 Swedish Grand Prix was the seventh round of the 1976 Formula One season and the ninth Swedish Grand Prix.`	`1976 Swedish Grand Prix point in time 13 June 1976, part of 1976 Formula One season.`	`1976 Swedish Grand Prix pole position Jody Scheckter, winner Jody Scheckter.`
`1976 Swedish Grand Prix was the seventh round of the 1976 Formula One season and the ninth Swedish Grand Prix.`	`1976 Swedish Grand Prix point in time 13 June 1976, part of 1976 Formula One season.`	`1976 Swedish Grand Prix point in time 13 June 1976, country Sweden.`
`1976 Swedish Grand Prix was the seventh round of the 1976 Formula One season and the ninth Swedish Grand Prix.`	`1976 Swedish Grand Prix point in time 13 June 1976, part of 1976 Formula One season.`	`1976 Swedish Grand Prix point in time 13 June 1976.`

Loss: CachedMultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

tekgen-ctsp

Dataset: tekgen-ctsp at 8d091eb
Size: 10,866 evaluation samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative
type	string	string	string
details	min: 13 tokens mean: 40.18 tokens max: 183 tokens	min: 7 tokens mean: 19.82 tokens max: 62 tokens	min: 7 tokens mean: 19.82 tokens max: 62 tokens

Samples:

anchor	positive	negative
`Two men with prior criminal records, Dieter Degowski and Hans-Jürgen Rösner, went on the run for two days through Germany and the Netherlands.`	`Gladbeck hostage crisis country Netherlands, country Germany, participant Hans-Jürgen Rösner, participant Dieter Degowski.`	`Gladbeck hostage crisis end time 18 August 1988, point in time 18 August 1988, country Germany, start time 16 August 1988.`
`The Gladbeck hostage crisis (known in Germany as the Gladbeck hostage drama) was a hostage-taking crisis that happened in August 1988 after an armed bank raid in Gladbeck, North Rhine-Westphalia, West Germany.`	`Gladbeck hostage crisis end time 18 August 1988, point in time 18 August 1988, country Germany, start time 16 August 1988.`	`Gladbeck hostage crisis country Netherlands, country Germany, participant Hans-Jürgen Rösner, participant Dieter Degowski.`
`The album was originally released only on cassette tape before later being made available for digital download on iTunes and similar digital media stores.`	`Vongole Fisarmonica instance of Album.`	`Vongole Fisarmonica performer Those Darn Accordions, publication date 01 January 1992, instance of Album.`

Loss: CachedMultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 512
per_device_eval_batch_size: 512
learning_rate: 8e-05
num_train_epochs: 1
warmup_ratio: 0.05
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: 512
per_device_eval_batch_size: 512
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: 8e-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: 1
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.05
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

Epoch	Step	Training Loss	YesaOuO/TEKGEN-CTSP_cosine_accuracy
-1	-1	-	0.6585
0.2252	500	0.6404	-
0.4505	1000	0.212	-
0.6757	1500	0.1764	-
0.9009	2000	0.1562	-
-1	-1	-	0.9166

Framework Versions

Python: 3.11.11
Sentence Transformers: 3.4.1
Transformers: 4.49.0
PyTorch: 2.6.0+cu124
Accelerate: 1.5.2
Datasets: 3.3.2
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",
}

CachedMultipleNegativesRankingLoss

@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}