wli14/HistAgent-Qwen3-8B-LoRA

HistAgent Qwen3-8B LoRA

This repository provides the LoRA adapter for the Qwen3-8B component of HistAgent. The model uses structured molecular and spatial evidence from a selected tissue spot to support evidence-grounded, multi-turn biological analysis.

Model details

• Base model: Qwen/Qwen3-8B
• Adapter: LoRA, r=16, alpha=32, dropout 0.05
• Target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
• Input: a spot-level evidence card and a biological question
• Output: a natural-language answer grounded in the supplied evidence

The adapter is used with thinking disabled so that only the final answer is returned.

Loading the adapter

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

base_model_id = "Qwen/Qwen3-8B"
adapter_id = "wli14/HistAgent-Qwen3-8B-LoRA"

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

Questions should be supplied with the selected spot's evidence card using the Qwen3 chat template and enable_thinking=False.

Intended use

This adapter supports the HistAgent interactive interface and research on evidence-grounded interpretation of spatial molecular readouts. It is not intended for clinical diagnosis or treatment decisions. Conclusions should be interpreted in the context of the supplied evidence and validated against measured data when appropriate.

Resources

• HistAgent interface: https://histagent.bio/histagent/
• GitHub: https://github.com/zipging/HistAgent
• HistAgent resources: https://huggingface.co/wli13/HistAgent