This repository contains a LoRA adapter for Qwen2.5-7B-Instruct fine-tuned for high-energy-physics question answering around signal signatures and Standard Model background composition.

Model Details

Base model: Qwen/Qwen2.5-7B-Instruct
Training method: supervised fine-tuning with LoRA
LoRA rank: 16
Checkpoint: qwen2_5_7b_lora16_sigbg_irred_red_v1, step 2200
Training code: ho22joshua/hep-posttraining
Dataset: ho22joshua/hep-signature-backgrounds

Intended Use

The model is intended for research workflows that ask for concise HEP reasoning about final-state signatures, dominant backgrounds, irreducible backgrounds, and reducible backgrounds.

Quick Start

python

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

base_model_id = "Qwen/Qwen2.5-7B-Instruct"
adapter_id = "ho22joshua/hep-qwen2.5-7b-lora16-sigbg-irred-red-step2200"

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

pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
messages = [
    {"role": "user", "content": "For a search targeting H to AA to photons, describe the signal signature and likely Standard Model backgrounds."}
]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
print(pipe(prompt, max_new_tokens=300, do_sample=False)[0]["generated_text"])

OpenAI-Compatible Local Endpoint

For endpoint serving, either load the adapter with a serving runtime that supports LoRA, or merge the adapter into the base model first.

bash

python -m vllm.entrypoints.openai.api_server \
  --model Qwen/Qwen2.5-7B-Instruct \
  --enable-lora \
  --lora-modules hep=ho22joshua/hep-qwen2.5-7b-lora16-sigbg-irred-red-step2200 \
  --host 0.0.0.0 \
  --port 8000 \
  --dtype bfloat16

bash

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "hep",
    "messages": [
      {"role": "user", "content": "For H to AA to photons, what backgrounds should be considered?"}
    ],
    "max_tokens": 300
  }'

Limitations

This model is a research artifact. It can make factual or physics mistakes and should not be used as a substitute for expert review, validated simulation, internal collaboration guidance, or publication-quality analysis approval.

License

The model license follows the constraints of the base model and the generated training data. Please verify those terms before downstream redistribution or commercial use.