ChaThai-gemma3-12b

Model Overview

ChaThai-gemma3-12b is a Thai language model fine-tuned from Google's Gemma-3-12B-IT base model, specifically optimized for medical domain applications in Thai. This model has been trained on medical data to provide accurate and contextually appropriate responses for healthcare-related queries in Thai.

Model Details

• Model Name: CMKL/ChaThai-gemma3-12b
• Base Model: google/gemma-3-12b-it
• Model Type: Causal Language Model
• Language: Thai (ไทย)
• Domain: Medical/Healthcare
• Parameters: 12 billion
• Architecture: Gemma-3

Training Data

The model has been fine-tuned on medical data including:

• Thai medical texts and documentation
• Medical terminology and concepts in Thai
• Healthcare-related conversations and Q&A pairs

Note: All training data has been properly anonymized and preprocessed to ensure patient privacy and compliance with medical data regulations.

Intended Use

Primary Use Cases

• Medical information assistance in Thai
• Healthcare chatbot applications
• Medical text generation and completion
• Medical terminology explanation

Limitations and Considerations

Medical Disclaimer

⚠️ Important: This model is for informational purposes only and should not be used as a substitute for professional medical advice, diagnosis, or treatment. Always consult with qualified healthcare professionals for medical decisions.

Known Limitations

• May occasionally generate medically incorrect information
• Limited to Thai language understanding and generation
• Performance may vary on rare medical conditions
• Requires human oversight for clinical applications
• May reflect biases present in training data

Ethical Considerations

• Designed to provide helpful medical information while emphasizing the importance of professional medical consultation
• Trained to avoid providing specific medical diagnoses or treatment recommendations
• Includes safety measures to redirect users to healthcare professionals when appropriate

Performance

Performance metrics and evaluation results would be included here based on your specific testing and validation.

Usage

# pip install accelerate

from transformers import AutoProcessor, Gemma3ForConditionalGeneration
from PIL import Image
import requests
import torch

model_id = "CMKL/ChaThai-gemma3-12b"

model = Gemma3ForConditionalGeneration.from_pretrained(
    model_id, device_map="auto"
).eval()

processor = AutoProcessor.from_pretrained(model_id)

messages = [
    {
        "role": "system",
        "content": [{"type": "text", "text": "You are a helpful assistant."}]
    },
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "อาการปวดหัวเป็นอย่างไร."}
        ]
    }
]

inputs = processor.apply_chat_template(
    messages, add_generation_prompt=True, tokenize=True,
    return_dict=True, return_tensors="pt"
).to(model.device, dtype=torch.bfloat16)

input_len = inputs["input_ids"].shape[-1]

with torch.inference_mode():
    generation = model.generate(**inputs, max_new_tokens=100, do_sample=False)
    generation = generation[0][input_len:]

decoded = processor.decode(generation, skip_special_tokens=True)
print(decoded)

Model Card Authors

CMKL MMFM Team

License

This model inherits the license from the base Gemma-3-12B-it model. Please refer to Google's Gemma license terms for usage rights and restrictions.

Contact

For questions, issues, or collaboration opportunities, please contact the CMKL team.

Acknowledgments

• Google AI for the Gemma-3-12B-it base model
• Medical data providers and healthcare institutions
• Thai language and medical domain experts who contributed to the training process

---

Disclaimer: This model is provided "as is" without warranty of any kind. Users are responsible for ensuring appropriate use in their specific applications and compliance with relevant regulations.