SentenceTransformer based on google/embeddinggemma-300m

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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("yasserrmd/dental-gemma-300m-emb")
# Run inference
queries = [
    "What is tooth transposition and how does it occur?\n",
]
documents = [
    'Tooth transposition is a condition where two adjacent teeth switch positions in the dental arch. It can be classified as complete transposition when both the crowns and roots of the involved teeth exchange places, or incomplete transposition when only the crowns are transposed. Tooth transposition is more commonly found unilaterally than bilaterally, with a higher prevalence in the maxillary (upper) arch. There is no sex preference for tooth transposition. The condition is not significantly related to other dental anomalies such as missing teeth, peg-shaped or hypoplastic teeth, or impacted teeth. The etiology of tooth transposition is believed to be genetically involved, and it most often occurs at the maxillary canine.',
    'Air polishing offers many advantages to clinicians and their patients. It is less time-consuming and effective in heavily stained surfaces, such as smoking and chlorhexidine stain. This minimizes operator and patient fatigue. Additionally, air polishing treatments can provide superior treatment outcomes in terms of speed and efficacy.',
    'The main forms of treatment for oral and tooth health in government hospitals and health centers in Turkey are extraction, removable prosthodontics, and routine amalgam and composite restorations. Root canal treatment is not generally preferred due to a lack of endodontic equipment and time.',
]
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([[0.7669, 0.0321, 0.2308]])

Training Details

Training Dataset

Unnamed Dataset

Size: 20,000 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: 9 tokens
mean: 19.66 tokens
max: 49 tokens

min: 22 tokens
mean: 87.28 tokens
max: 221 tokens

	sentence_0	sentence_1
type	string	string
details	min: 9 tokens mean: 19.66 tokens max: 49 tokens	min: 22 tokens mean: 87.28 tokens max: 221 tokens

Samples:

sentence_0	sentence_1
`How do prosthodontists in Australia determine the need for placing a post in a tooth restoration?`	`Prosthodontists in Australia consider both the quantity of tooth structure and the type of planned restoration when deciding whether to place a post. The location of the tooth in the arch seems to have less influence on their decision. Molar teeth and mandibular anterior teeth are less likely to receive posts.`
`What are some patient-centered outcome measures used to assess the impact of oral health problems on quality of life?`	There are several patient-centered outcome measures called 'oral health related quality of life measures' (OHQoL) that have been developed to assess the extent to which oral health problems affect a person's quality of life. Two measures that have received particular attention are the Oral Health Impact Profile (OHIP-14) and the UK Oral Health Related Quality of Life (OHQoL-UK) questionnaires. The OHIP-14 measures the adverse impacts of oral conditions on daily life, while the OHQoL-UK incorporates both negative and positive influences on health.
`How does finite element analysis (FEA) contribute to the understanding and improvement of dental implant procedures?`	Finite element analysis (FEA) is a computational method that can be used to simulate the distribution of stress and strain in the mandibular bone and osseointegrated implants. By considering various variables such as material characteristics, types of loads, and individual bio-subjectivity, FEA studies provide valuable insights into stress distribution and geometry evaluation. This information helps in making informed decisions about implant positioning, inclination, and type to ensure the long-term stability and success of dental implants.

Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 4
per_device_eval_batch_size: 4
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: 4
per_device_eval_batch_size: 4
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: {}
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}
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
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: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.1	500	0.0147
0.2	1000	0.0142
0.3	1500	0.0154
0.4	2000	0.0085
0.5	2500	0.0052
0.6	3000	0.0071
0.7	3500	0.0025
0.8	4000	0.0028
0.9	4500	0.0056
1.0	5000	0.0045

Framework Versions

Python: 3.12.11
Sentence Transformers: 5.1.0
Transformers: 4.56.1
PyTorch: 2.8.0+cu128
Accelerate: 1.10.1
Datasets: 4.0.0
Tokenizers: 0.22.0

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",
}

MultipleNegativesRankingLoss

@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}
}