Med vLLM (Config-first Repository)

This repository serves as a config-first landing for the Med vLLM stack.

It contains example configuration files and is intended to help users discover and consume the MedicalModelConfig from the Hub via from_pretrained, and to use these as starting points for training or inference in medical NLP tasks.

Contents

  • NER config example (examples/ner/)
  • Classification config example (examples/classification/)
  • Generation config example (examples/generation/)

Install

pip install medvllm

Quickstart (Python)

from medvllm.medical.config.models.medical_config import MedicalModelConfig
cfg = MedicalModelConfig.from_pretrained("Junaidi-AI/med-vllm")
print(cfg.task_type)

Or directly load a specific example folder if exported as a subfolder with its own config files.

Examples

Use these as starting points and customize fields like task_type, classification_labels, medical_entity_types, and domain settings.

Tasks supported

  • Named Entity Recognition (NER)
  • Text Classification
  • Text Generation

All tasks share a unified configuration schema via MedicalModelConfig.

Weights roadmap

This repo currently focuses on configs. Model weights/adapters will be added progressively:

  • BioBERT/ClinicalBERT adapters
  • Task-specific fine-tuned checkpoints (NER/Classification)

Follow the repo for updates or open a Discussion to request specific checkpoints.

Debug and logging

By default, verbose config debug prints are silenced. To enable them for troubleshooting, set:

export MEDVLLM_CONFIG_DEBUG=1

Medical Disclaimer

This repository and associated configurations are provided for research and engineering purposes only. They are not intended for clinical decision-making. Always involve qualified healthcare professionals and ensure compliance with applicable regulations (e.g., HIPAA, GDPR). Avoid using PHI/PII.

License

MIT

Downloads last month
36
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Spaces using Junaidi-AI/med-vllm 2