ChenyuEcho/hospital_emaillevel_oldtrainmethod

SentenceTransformer based on Qwen/Qwen3-Embedding-0.6B

This is a sentence-transformers model finetuned from Qwen/Qwen3-Embedding-0.6B. 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: Qwen/Qwen3-Embedding-0.6B
• Maximum Sequence Length: 768 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': 768, 'do_lower_case': False, 'architecture': 'PeftModelForFeatureExtraction'})
  (1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': True, '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
queries = [
    "What is the proposed date/time for a meeting with Maintenance to review findings and strategize a remediation plan for OR wing HVAC fluctuations affecting post-op mobility protocols?",
]
documents = [
    'Subject: HVAC System Issues Impacting OR Wing: Request for Collaborative Solution\nFrom: Dr. Susan L. Chang\nTo: Brian K. Lee\nDate: 2025-10-20\n\nHi Brian,\n\nThank you for raising this concern regarding the HVAC fluctuations in the OR wing. My team has indeed noticed some difficulties in maintaining prescribed post-operative mobility protocols due to variable temperatures, which can discourage early patient movement and affect our recovery benchmarks. I agree that partnering with Maintenance for timely remediation is essential; I can share specific observations from our unit that might help guide targeted improvements in the HVAC settings or scheduling.\n\nLet me know when would be best to meet with you and Maintenance so we can review findings and strategize a plan. Thanks again for the proactive outreach.\n\nBest,\nSusan',
    "Subject: Payment Posting Discrepancy – Assistance Needed with Discharge Billing\nFrom: Isaiah T. Jackson\nTo: Taylor A. Richardson\nDate: 2025-10-30\n\nHi Taylor,\n\nThanks for bringing this to my attention. I've pulled the discharge plan and placement records and will cross-reference them with the insurance authorization to verify alignment. Once I have more clarity, I’ll coordinate with you and finance as needed to resolve any discrepancies—please hold off on correction for now. If I require any additional documentation, I’ll let you know.\n\nBest,\nIsaiah",
    "Subject: Request for Conference Call Number\nFrom: Carol A. Campbell\nTo: Melissa K. King\nDate: 2025-10-27\n\nHi Melissa,\n\nI hope you're doing well. I am reaching out to request the conference call number for our upcoming credentialing review meeting scheduled later this week. Having the dial-in details ahead of time will help me circulate the information to all committee members and ensure we are fully prepared to discuss the physician files on the agenda. Please let me know at your earliest convenience if there are any specific protocols or security codes required for access.\n\nThank you very much for your assistance.\n\nBest regards,\nCarol",
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 1024] [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.7461, 0.0432, 0.0610]], dtype=torch.bfloat16)

Evaluation

Metrics

Information Retrieval

• Dataset: val_evaluation
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.8948
cosine_accuracy@3	0.9583
cosine_accuracy@5	0.98
cosine_accuracy@10	1.0
cosine_precision@1	0.8948
cosine_precision@3	0.3194
cosine_precision@5	0.196
cosine_precision@10	0.1
cosine_recall@1	0.8948
cosine_recall@3	0.9583
cosine_recall@5	0.98
cosine_recall@10	1.0
cosine_ndcg@10	0.9472
cosine_mrr@10	0.9302
cosine_map@100	0.9302

Training Details

Training Dataset

Unnamed Dataset

• Size: 2,394 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: 8 tokens mean: 26.77 tokens max: 63 tokens	min: 99 tokens mean: 160.64 tokens max: 399 tokens

• Samples:

  sentence_0	sentence_1
  What are the routine badge access needs for the phlebotomy team under the updated infection control protocol to prevent access-related delays in the patient observation wing?	Subject: Request for Assistance: Infection Control Protocol Update Impacting Phlebotomy Access From: Paul R. Nelson To: Carlos J. Rodriguez Date: 2025-12-15 Hi Carlos, I wanted to reach out regarding recent changes to our infection control protocols, which now require stricter badge access in certain areas, including the patient observation wing. We've had a few incidents where phlebotomy staff were delayed due to these new access restrictions. To ensure patient care isn't impacted, could you assist by identifying routine access needs for your team so we can adjust permissions accordingly? Please let me know if you or your team have experienced any specific challenges, and I’ll work to resolve these promptly. Thanks for your cooperation, Paul
  Are there any specific legal/compliance issues to be addressed during the Emergency Preparedness Drill?	Subject: Re: Upcoming Hospital-wide Emergency Preparedness Drill – Participation Required From: David R. Park To: Richard T. Howard Date: 2025-12-02 Hello Richard, Thank you for the detailed announcement and for outlining the objectives of the upcoming Emergency Preparedness Drill. I appreciate the proactive approach to ensuring all staff are familiar with emergency protocols, especially with the simulation of both power and network outages. My team and I will review our current department safety procedures and ensure we are prepared to both participate in the drill and report any issues we discover. Please let me know if there are any specific compliance issues from the legal perspective you would like addressed during the exercise. Best regards, David R. Park
  Quem é o responsável por fornecer o relatório de ultrassom de Miguel Silva para desbloquear a baixa de dívida incobrável?	Subject: Solicitação de apoio: aprovação de baixa de dívidas incobráveis From: Rachel K. Martinez To: Jasmine K. Patel Date: 2025-10-27 Olá Jasmine, Estou enfrentando dificuldades com a aprovação de uma baixa de dívida incobrável referente ao paciente do Pronto-Socorro. O sistema está bloqueando o processo devido à ausência de alguns documentos do setor de imagem. Você poderia, por favor, me ajudar a localizar ou encaminhar o relatório do ultrassom do paciente Miguel Silva? Precisamos desse documento para dar andamento à solicitação. Agradeço pela agilidade! Atenciosamente, Rachel K. Martinez

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 8
• per_device_eval_batch_size: 8
• 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: 3
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_ratio: None
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• enable_jit_checkpoint: False
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• use_cpu: False
• seed: 42
• data_seed: None
• bf16: False
• fp16: False
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: -1
• ddp_backend: None
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• 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
• 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
• 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_for_metrics: []
• eval_do_concat_batches: True
• auto_find_batch_size: False
• full_determinism: False
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• 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
• use_cache: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	val_evaluation_cosine_ndcg@10
1.0	300	-	0.9461
1.6667	500	0.0168	-
2.0	600	-	0.9459
3.0	900	-	0.9472

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.2.2
• Transformers: 5.0.0
• PyTorch: 2.9.0+cu128
• Accelerate: 1.12.0
• Datasets: 4.0.0
• 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",
}

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}
}