yasserrmd/ophthalmology-gemma-300m-emb

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/ophthalmology-gemma-300m-emb")
# Run inference
queries = [
    "How does intravitreal injection of tacrolimus affect the expression of chemokines in the retina?\n",
]
documents = [
    'Intravitreal injection of tacrolimus has been shown to markedly reduce the expression of monocyte chemoattractant protein-1 (MCP-1) and regulated upon activation, normal T cell expressed and secreted (RANTES) mRNA in the retina of immunized rats. MCP-1 and RANTES are potent chemoattractants for T lymphocytes and macrophages, which are infiltrating cells observed in the posterior segment of eyes with ocular inflammation. The suppression of MCP-1 and RANTES gene expression by tacrolimus may lead to a reduced number of infiltrating T lymphocytes and macrophages in the vitreous cavity and retina, thereby reducing the inflammatory reactions associated with ocular inflammation.',
    'Anaplastic OAs are characterized by intensive vascularity. The choice of laser radiation for tumor removal depends on the vascularity observed. For low-grade OAs with minimal vascularity, laser vaporization using radiation with a wavelength of 1.47 μm and power of 7 W is planned. However, for anaplastic OAs with intense vascularity, semiconductor laser radiation with a wavelength of 0.808 μm and power of 30 W is used due to its good coagulating properties.',
    'For patients with severe eye changes and diplopia, high dose prednisone (120 mg/day) can be effective in reducing symptoms. Some clinicians also add azathioprine as a steroid sparing agent during long term treatment. Orbital decompression operations and a combination of orbital radiotherapy and steroids have also shown to be effective in reducing proptosis and diplopia in some patients. Corrective muscle surgery may be undertaken for residual diplopia when the eyes are quiet.',
]
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([[0.6629, 0.0904, 0.0715]])

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: 10 tokens mean: 21.87 tokens max: 67 tokens	min: 15 tokens mean: 90.2 tokens max: 241 tokens

• Samples:

  sentence_0	sentence_1
  How does OCT contribute to the identification and assessment of vitreoretinal traction in macular holes, and why is bilaterally performing tomographic imaging important for early detection in patients affected by this condition?	OCT contributes significantly to the identification and assessment of vitreoretinal traction in macular holes, enabling the visualization of vitreoretinal interface abnormalities. It plays a crucial role in assessing the risk for hole formation in the fellow eye of patients with a unilateral macular hole, as the probability of developing a macular hole in the contralateral eye is 13% in 48 months. Therefore, it is essential to perform bilateral tomographic imaging in patients affected by this pathology for early detection in the other eye. Considering that surgery in the normal contralateral eye of patients with macular holes is not acceptable, early detection through bilateral imaging is crucial for appropriate clinical management.
  How does corneal refractive therapy, also known as overnight orthokeratology, work to reduce myopia?	Corneal refractive therapy, or overnight orthokeratology, involves the use of reverse geometry contact lenses to induce temporary flattening in the central corneal curvature. This temporary flattening leads to a short-term reduction in myopia and improved unaided visual acuity. Recent studies have also shown that corneal refractive therapy has the potential to slow down the progression of myopia.
  What is the significance of contrast sensitivity testing in early age-related macular degeneration (AMD) patients, and how does it relate to the prognosis and treatment of the condition?	Contrast sensitivity testing is crucial in early AMD patients as it can detect even slight changes in visual system performance, particularly impairment at nighttime with and without glare. This testing provides important prognostic value, as AMD patients with less affected or intact contrast sensitivity have better prognoses and respond more effectively to treatments with vascular endothelial growth factor inhibitors, photodynamic therapy, and laser photocoagulation. Additionally, early diagnosis of AMD using contrast sensitivity testing can aid in the prevention of future blindness.

• 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	0.0239
0.2999	1000	0.0192
0.4499	1500	0.015
0.5999	2000	0.0058
0.7499	2500	0.0071
0.8998	3000	0.0072

Framework Versions

• Python: 3.12.11
• Sentence Transformers: 5.1.0
• Transformers: 4.56.2
• PyTorch: 2.8.0+cu128
• Accelerate: 1.10.1
• Datasets: 4.0.0
• Tokenizers: 0.22.1

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",
}

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}
}