TxGemma-9B-LiPo (LoRA Adapter)

LoRA adapter for TxGemma-9B-Chat finetuned on a lipid nanoparticle (LNP) mRNA transfection efficiency dataset. From the ACL 2026 paper:

LipoAgent: Coordinating Fine-Tuned LLM Agents for Safer Lipid Design

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

base_model_id = "google/txgemma-9b-chat"
adapter_id = "Nemo0412/txgemma-9b-lipo"

tokenizer = AutoTokenizer.from_pretrained(base_model_id)
base_model = AutoModelForCausalLM.from_pretrained(
    base_model_id, torch_dtype=torch.bfloat16, device_map="auto"
)
model = PeftModel.from_pretrained(base_model, adapter_id)

Training Details

• Base model: google/txgemma-9b-chat
• Method: LoRA (r=16, α=32)
• Dataset: ~12,000 LNP samples with mRNA transfection efficiency labels
• Epochs: 3 | Train loss: ~0.21 | Eval loss: 0.261
• Hardware: NVIDIA H100 80GB