This is a sentence-transformers model finetuned from google/embeddinggemma-300m. 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/embeddinggemma-300m
Maximum Sequence Length: 2048 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': 2048, 'do_lower_case': False, 'architecture': 'Gemma3TextModel'})
  (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): Dense({'in_features': 768, 'out_features': 3072, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity'})
  (3): Dense({'in_features': 3072, 'out_features': 768, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity'})
  (4): 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("KhanhDao1512/my-embedding-gemma")
# Run inference
queries = [
    "Body: fingerblasting implies it being consentual\n\nmore lol i raped my own sister lol xd randum haha shell always be haunted by what i did to her lmao | Rule: no legal advice do not offer or request legal advice | Violation: no violation",
]
documents = [
    'gather evidence see a lawyer immediately preemptive strike',
    'a quick google search found this you can trick the card into thinking it’s doing a chipandsignature transaction while the terminal thinks it’s chipandpin the upshot is that you can buy stuff using a stolen card and a pin of 0000 or anything you want\n\nsource securityaffairs co wordpress 25134 cyber crime chip and pin cloning html',
    'get 50 discount on all academic writing and editing services at cheapcustomwriting\n\n\nplagiarism free assignment writing essay writing dissertation writing thesis writing phd proposal writing research paper writing coursework writing medical writing term paper writing and resume writing at cheapcustomwriting\n\nour active 50 code is nbxqwruf6ewv\n\nvisit wwwcheapcustomwritingcom',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 768] [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.9996, 0.3074, 0.1825]])

Training Details

Training Dataset

Unnamed Dataset

Size: 8,116 training samples
Columns: anchor, positive, and negative
Approximate statistics based on the first 1000 samples:
anchor positive negative
type string string string
details
min: 32 tokens
mean: 60.21 tokens
max: 132 tokens
min: 11 tokens
mean: 40.74 tokens
max: 139 tokens
min: 6 tokens
mean: 33.66 tokens
max: 162 tokens

Samples:

anchor	positive	negative
`Body: banks dont want you to know this click here to know more \| Rule: no advertising spam referral links unsolicited advertising and promotional content are not allowed \| Violation: no violation`	`if you could tell your younger self something different about sex what would that be[object Object][object Object]i am in a contest to win funding for my sex positive film vote here[object Object][object Object]sheknows offerpop campaign 813112 entry v144417`	`watch golden globe awards 2017 live online in hd coverage without ads vip streams[object Object][object Object][object Object]hd stream quality watch link1forum submitexpress viewtopic php[object Object][object Object][object Object]hd broadcasting quality watch link1forum submitexpress viewtopic php[object Object][object Object][object Object]mobile compatibility yes[object Object][object Object][object Object]no ads no ads ads[object Object][object Object]`
`Body: banks dont want you to know this click here to know more \| Rule: no advertising spam referral links unsolicited advertising and promotional content are not allowed \| Violation: no violation`	`if you could tell your younger self something different about sex what would that be[object Object][object Object]i am in a contest to win funding for my sex positive film vote here[object Object][object Object]sheknows offerpop campaign 813112 entry v144417`	`double cee x bands epps birds[object Object][object Object]downloadstream[object Object][object Object]music theblacksmithed download birds`
`Body: banks dont want you to know this click here to know more \| Rule: no advertising spam referral links unsolicited advertising and promotional content are not allowed \| Violation: no violation`	`hunt for lady for jack off in neighbourhood url inmusi gakq`	`watch golden globe awards 2017 live online in hd coverage without ads vip streams[object Object][object Object][object Object]hd stream quality watch link1forum submitexpress viewtopic php[object Object][object Object][object Object]hd broadcasting quality watch link1forum submitexpress viewtopic php[object Object][object Object][object Object]mobile compatibility yes[object Object][object Object][object Object]no ads no ads ads[object Object][object Object]`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim", 4 "gather_across_devices": false 5}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 1
learning_rate: 2e-05
warmup_ratio: 0.1
prompts: task: sentence similarity | query:

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 1
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: 2e-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: 3
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
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}
parallelism_config: 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
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: False
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: False
prompts: task: sentence similarity | query:
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
2.0	8116	0.5455

Framework Versions

Python: 3.11.13
Sentence Transformers: 5.1.0
Transformers: 4.57.0.dev0
PyTorch: 2.6.0+cu124
Accelerate: 1.8.1
Datasets: 3.6.0
Tokenizers: 0.22.0

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",
}

MultipleNegativesRankingLoss

bibtex

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

	anchor	positive	negative
type	string	string	string
details	min: 32 tokens mean: 60.21 tokens max: 132 tokens	min: 11 tokens mean: 40.74 tokens max: 139 tokens	min: 6 tokens mean: 33.66 tokens max: 162 tokens