metadata
license: apache-2.0
tags:
- pharmacy
- clinical-decision-support
- medication-review
- mistral
- fine-tuned
base_model: mistralai/Mistral-7B-Instruct-v0.1
model_name: markoste-clinical-mistral
language: en
datasets:
- Markolo96/Mistral_DB
💊 Markoste Clinical Mistral (7B)
A pharmacist-optimized large language model fine-tuned on Australian Product Information (PI) Q&A data for clinical medication review support.
🧠 Model Purpose
Markoste Clinical Mistral is fine-tuned to answer medication-specific clinical questions such as:
- Dosing adjustments in renal impairment
- Contraindications
- IV administration instructions
- Interactions and precautions
- Pharmacokinetic properties
- Pregnancy & breastfeeding guidance
🧪 Training
Fine-tuned using LoRA on top of mistralai/Mistral-7B-Instruct-v0.1 with Q&A pairs derived from structured TGA PI documents.
⚙️ Example Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("Markolo96/markoste-clinical-mistral", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("Markolo96/markoste-clinical-mistral")
prompt = "<|user|>\nWhat are the indications for erythromycin lactobionate?\n<|assistant|>"
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))