SentenceTransformer based on lightonai/modernbert-embed-large

This is a sentence-transformers model finetuned from lightonai/modernbert-embed-large. It maps sentences & paragraphs to a 1024-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: lightonai/modernbert-embed-large
Maximum Sequence Length: 256 tokens
Output Dimensionality: 1024 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': 256, 'do_lower_case': False, 'architecture': 'ModernBertModel'})
  (1): Pooling({'word_embedding_dimension': 1024, '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 = [
    'Instruct: Retrieve semantically similar text based on safety profile\nQuery: We will see women as popes  when Islamic men stop raping women, playing tahharush, killing gays, forcing them to cover up, not allowing them to drive, stop performing genital mutilation etc. Jk we know neither will ever happen. Now run along and make my white self a sandwich honey!!',
    'Instruct: Retrieve semantically similar text based on safety profile\nQuery: It\'s official. Canadians have been driven to mental illness. \n\n(The test for civility here is confusing. I rated a comment referring to the left as disgusting and vile as not good but civil. There were no personal names. "Vile" means disgusted, and the poster was disgusted. So ??? If that\'s not OK, why is "Fiberals" allowed? That\'s referring to the Liberals as liars.)',
    "Instruct: Retrieve semantically similar text based on safety profile\nQuery: Right, so instead of proving your claim about the Hawaiian haters in this thread, you instead respond with the typical racist Hawaiian stereotypes about haoles.\n\nI see racism in Hawaii because ignorant racists like you pretend that your blind hatred is something other than what it is.\n\nThe fact is I have said nothing different than the kanaka maoil who were arrested outside of the Na'i Aupuni convention, but the difference is, you can not stand hearing it from someone who you assume does not have the magic blood.\n\nNews flash. Every single person on this planet is indigenous. Get over yourself.",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.9453, 0.9609],
#         [0.9453, 1.0000, 0.9609],
#         [0.9609, 0.9609, 1.0000]], dtype=torch.bfloat16)

Training Details

Training Dataset

Unnamed Dataset

Size: 12,868 training samples
Columns: sentence_0, sentence_1, and label
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1 label
type string string float
details
min: 20 tokens
mean: 82.03 tokens
max: 256 tokens

min: 20 tokens
mean: 84.69 tokens
max: 256 tokens

min: 0.0
mean: 0.51
max: 1.0

	sentence_0	sentence_1	label
type	string	string	float
details	min: 20 tokens mean: 82.03 tokens max: 256 tokens	min: 20 tokens mean: 84.69 tokens max: 256 tokens	min: 0.0 mean: 0.51 max: 1.0

Samples:

sentence_0	sentence_1	label
Instruct: Retrieve semantically similar text based on safety profile Query: You threatened to kill people when you made this statement. "Except maybe criminals who have to consider whether their next victim is packing heat. Which by the way, I don't. Probably." Too many of you white folks refuse to take responsibility for your choices. I consider the GOP, in the 21st century, to be little more than a white identity death cult. You love being white and armed more than being human beings or Americans. You have made it clear that you have no problem taking a human life. Do you support the state sponsored killing of unarmed black people? BTW, you didn't respond to my comments about the Vietnam war. Are you not brave?	`Instruct: Retrieve semantically similar text based on safety profile Query: Shoot the messenger--an ancient tradition.`	`1.0`
Instruct: Retrieve semantically similar text based on safety profile Query: This land is OUR land, David. If Seneca wants to do logging, then do it on private land, the problem is they have over cut their private lands and now they want to get their greedy hands on the last of OUR public forests. I have a job, I have been working since I was eleven years old and am nearing retirement age. I have also been a volunteer for over 25 years working to help us transition our economy and consciousness to one that is truly sustainable. We are in the midst of the most devastating crisis humans have ever faced due to burning fossil fuels, deforestation, over fishing, pollution, chemical farming and over human population. We have created the sixth mass species extinction! Wake up, David and the rest of you who think working to protect our last wildlands- and the biosphere in general- is "wacko." What is "wacko" is those of you who are in denial of this global crisis. Face facts.	`Instruct: Retrieve semantically similar text based on safety profile Query: Well that's a new level of ineptitude. They literally didn't get their own memo. I'm sure it was entirely accidental...`	`1.0`
`Instruct: Retrieve semantically similar text based on safety profile Query: What's the matter? Did he get too close to the truth for you by characterizing Trump as a rich narcissist, materialist, demagogue who doesn't really have any coherent ''program'' other than staying at the center of attention, expanding his family's wealth and influence, and being applauded by his core constituency?`	`Instruct: Retrieve semantically similar text based on safety profile Query: He can get a translator. Trudeau just speaks nonsense in both.`	`1.0`

Loss: main.OptimizedContrastiveDistillationLoss

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 1
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: 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
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: 1
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: None
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
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}
parallelism_config: None
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch_fused
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
project: huggingface
trackio_space_id: trackio
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
hub_revision: None
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: no
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: True
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.6211	500	0.0186

Framework Versions

Python: 3.14.4
Sentence Transformers: 5.1.0
Transformers: 4.57.6
PyTorch: 2.11.0+cu128
Accelerate: 1.13.0
Datasets: 4.8.4
Tokenizers: 0.22.2

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",
}