Model Card for IOTAIRx/medgemma-27b-text-ft
Model Details
Model Description
MedGemma-3-27B-FT is a version of the google/medgemma-27b-text-it that has been fine-tuned on various medical datasets. Its goal is to improve performance on medical / clinical reasoning, question-answering, and domain knowledge tasks.
The base model is capable of medical text understanding and reasoning, and your fine-tuned version specializes it further toward the clinical knowledge domain.
- Developed by: IOTAIRx
- Model type: Language Model
- Language(s) (NLP): en
- License: mit
- Finetuned from model: google/medgemma-27b-text-it
Model Sources [optional]
Uses
Direct Use
This model is intended for research purposes in the medical and clinical domain. It can be used for tasks like question answering, summarization, and information extraction from clinical texts.
Downstream Use [optional]
The model can be further fine-tuned on more specific medical tasks or datasets to improve performance in specialized domains.
Out-of-Scope Use
This model is not approved for direct use in clinical care or diagnosis without validation and oversight. It may produce hallucinations or incorrect medical advice.
Bias, Risks, and Limitations
The model's performance is highly dependent on the quality and scope of the training data. Biases or gaps in the training data may be reflected in the model's outputs.
Recommendations
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Human verification of the model's output is crucial in any application.
How to Get Started with the Model
Use the code below to get started with the model.
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("IOTAIRx/medgemma-27b-text-ft")
model = AutoModelForCausalLM.from_pretrained("IOTAIRx/medgemma-27b-text-ft")
# Example inference
inputs = tokenizer("What is the normal range of hemoglobin in adults?", return_tensors="pt")
outputs = model.generate(**inputs)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
Training Details
Training Data
The model was fine-tuned on the https://huggingface.co/datasets/miriad/miriad-5.8M dataset and other proprietory clinical sources.
Training Procedure
Preprocessing [optional]
[More Information Needed]
Training Hyperparameters
- Training regime: bf16 mixed precision
Speeds, Sizes, Times [optional]
[More Information Needed]
Evaluation
Testing Data, Factors & Metrics
Testing Data
The model was evaluated on the test split of the cais/mmlu dataset, clinical_knowledge subset.
Factors
[More Information Needed]
Metrics
The primary metric used for evaluation was accuracy.
Results
Summary
- Dataset: mmlu_clinical_knowledge
- Accuracy: 0.834
- Standard Error: 0.0229
Environmental Impact
- Hardware Type: NVIDIA A100-SXM4-40GB
- Hours used: [More Information Needed]
- Cloud Provider: [More Information Needed]
- Compute Region: [More Information Needed]
- Carbon Emitted: [More Information Needed]
Technical Specifications [optional]
Model Architecture and Objective
The model is based on the Gemma 3 architecture, a transformer-based language model.
Compute Infrastructure
Hardware
- GPU: NVIDIA A100-SXM4-40GB
Software
- PyTorch version: 2.6.0+cu124
- transformers version: 4.51.3
- lm_eval version: 0.4.9.1
Citation [optional]
BibTeX:
[More Information Needed]
APA:
[More Information Needed]
- Downloads last month
- 34