silas114514/PMTX1-2B

Prompt Evolution Qwen3.5-2B Adapter (train_aligned)

This repository provides a LoRA adapter fine-tuned from Qwen/Qwen3.5-2B for Chinese prompt-rewriting coaching (train_aligned style).

What this is

• Format: PEFT LoRA adapter (adapter_model.safetensors)
• Base model required: Qwen/Qwen3.5-2B
• Not GGUF / not Ollama one-click by default

Quick Start (Transformers + PEFT)

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

base_model = "Qwen/Qwen3.5-2B"
adapter_repo = "<your-username>/prompt-evolution-qwen35-2b-adapter"

bnb = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.float16,
    bnb_4bit_use_double_quant=True,
)

tok = AutoTokenizer.from_pretrained(base_model, trust_remote_code=True)
if tok.pad_token is None:
    tok.pad_token = tok.eos_token or tok.unk_token

model = AutoModelForCausalLM.from_pretrained(
    base_model,
    quantization_config=bnb,
    torch_dtype=torch.float16,
    device_map="auto",
    trust_remote_code=True,
)
model = PeftModel.from_pretrained(model, adapter_repo)
model.eval()

prompt = "你是 Prompt Evolution 的提示词纠偏教练。请只做提示词优化，不要直接代做任务。\n原始提示词：写周报，你看着办就行，快一点。"
inputs = tok(prompt, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=220, do_sample=False)
print(tok.decode(out[0], skip_special_tokens=True))

Training Snapshot

• Prompt style: train_aligned
• Steps: 100
• Selected layers: top 8 transformer layers
• LoRA targets: q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj
• See run_summary.json, run_config.json, metrics_summary.json for details.

Notes

• Follow base model license/usage terms from Qwen/Qwen3.5-2B.
• This adapter is intended for prompt-rewriting assistance, not fact injection or task substitution.