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/neurology-gemma-300m-emb")
# Run inference
queries = [
    "How is the proportion of ischemic strokes treatable by endovascular thrombectomy (ET) within 8 hours of onset (ETT8) calculated?\n",
]
documents = [
    'The proportion of ischemic strokes treatable by endovascular thrombectomy (ET) within 8 hours of onset (ETT8) is calculated using the fraction of all AIS due to LVO (FLVO), the proportion of AIS with ED arrival within 6 hours of onset (ED6), the proportion of patients with stroke onset within 60 minutes of an endovascular center (AMB60), and the proportion of LVO patients arriving within 7 hours without any contraindications to ET (NoCONTR). The formula used is ETT8 = FLVO × ED6 × AMB60 × NoCONTR.',
    'At 3 months post-injury, patients with mild closed head injuries (mCHI) commonly experience symptoms such as headaches, dizziness, noise sensitivity, fatigue, poor memory, poor concentration, slowed thinking, and double vision. These symptoms were reported to be at least a mild problem for a significant number of patients with mCHI.',
    'Soluble Ab oligomers have been found to be more neurotoxic than Ab fibrils. Studies have shown that synapse loss was strongly correlated with the level of soluble Ab oligomers, but not monomeric Ab, insoluble fibrils, or APP levels. This suggests that the presence of soluble Ab oligomers has a significant impact on neuronal injury and synaptic activity.',
]
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([[ 0.6046, -0.0253, -0.0091]])

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: 21.87 tokens
max: 61 tokens

min: 21 tokens
mean: 92.13 tokens
max: 204 tokens

	sentence_0	sentence_1
type	string	string
details	min: 9 tokens mean: 21.87 tokens max: 61 tokens	min: 21 tokens mean: 92.13 tokens max: 204 tokens

Samples:

sentence_0	sentence_1
`What are the complications that can arise from the injection of the sphenopalatine ganglion?`	`Complications that can arise from the injection of the sphenopalatine ganglion include ecchymosis in the cellular tissue below the eye. While initially, there was only one reported case of complication, subsequent cases have shown that difficulties can arise following the injections. However, if the technique is carefully followed, serious complications can be avoided.`
`How does the interaction between CAPON and nNOS contribute to neuronal death after an excitotoxic stimulus?`	`CAPON, a protein specifically associated with nNOS, interacts with the PdZ domain of nNOS. The interaction between CAPON and nNOS mediates the nNOS-p38MAPK pathway during neuronal death after an excitotoxic stimulus. L-TAT-GESV, a cell-permeable peptide, can compete with CAPON for the unique PdZ domain of nNOS, leading to increased survival tissue in a severe model of neonatal hypoxia-ischemia. CAPON may serve as a high-specificity target for ischemia.`
`What are the motor signs associated with Parkinson's disease?`	`Parkinson's disease is characterized by motor signs such as hypokinesia (reduced movement) and abnormal posture. These motor signs are commonly observed in patients with Parkinson's disease.`

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: 6
per_device_eval_batch_size: 6
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: 6
per_device_eval_batch_size: 6
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
0.1500	500	0.026
0.2999	1000	0.0117
0.4499	1500	0.0068
0.5999	2000	0.0052
0.7499	2500	0.0047
0.8998	3000	0.0017

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}
}