Fine-tuned Gemma 3 1B for clinical and biomedical NLP.
BioGemma is a domain-adapted version of Google's Gemma 3 1B language model, fine-tuned on biomedical and clinical text. Target tasks include:
- Medical question answering
- Clinical text comprehension and summarization
- Biomedical entity recognition (drugs, conditions, procedures)
- Medical literature summarization
| Detail | Value |
|---|---|
| Base model | Google Gemma 3 1B |
| Method | LoRA / QLoRA (parameter-efficient fine-tuning) |
| Training data | Curated medical corpus — PubMed abstracts, medical textbooks, clinical guidelines |
| Hardware | Single-GPU training |
| Technology | Purpose | |
|---|---|---|
| 🧠 | Gemma 3 1B | Base language model |
| 🔥 | PyTorch | Deep learning framework |
| 🤗 | Transformers | Model loading and inference |
| 🚀 | Accelerate | Multi-device model distribution |
| 🎯 | PEFT / LoRA | Parameter-efficient fine-tuning |
pip install torch transformers accelerateimport torch
from transformers import AutoTokenizer, AutoModelForCausalLM
model_name = "stabgan/biogemma"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
dtype=torch.bfloat16,
device_map="auto",
)
prompt = "What are the common symptoms of acute kidney injury?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))- No model weights published yet. The HuggingFace Hub model (
stabgan/biogemma) is not available as of this writing. The usage example above will fail until weights are uploaded. - No training or evaluation code in this repository. The repo currently contains only documentation. There are no scripts, notebooks, configs, or reproducible training pipelines.
- No benchmark results. No evaluation against standard medical NLP benchmarks (MedQA, PubMedQA, MedMCQA) has been published.
- Publish model weights to HuggingFace Hub
- Add training and evaluation scripts
- Benchmark against MedQA, PubMedQA, MedMCQA
- Clinical note generation capabilities
- RLHF alignment for medical safety
This model is for research purposes only. It is not intended for clinical decision-making or medical diagnosis. Always consult qualified healthcare professionals for medical advice.