"""Merge a saved LoRA adapter into the base model, laid out like train_lora.py --merge-to does.
usage: merge.py <base model> <adapter dir> <out dir>"""
import os
import shutil
import sys

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

base, adapter, out = sys.argv[1:4]
dev = "cuda" if torch.cuda.is_available() else "cpu"
model = AutoModelForImageTextToText.from_pretrained(base, dtype=torch.bfloat16).to(dev)
model = PeftModel.from_pretrained(model, adapter)
merged = model.merge_and_unload()
merged.save_pretrained(out, safe_serialization=True)
AutoTokenizer.from_pretrained(base).save_pretrained(out)
src = base if os.path.isdir(base) else None
if src is None:
    from huggingface_hub import snapshot_download
    src = snapshot_download(base, allow_patterns=["*.json", "*.jinja", "*.txt"])
for f in os.listdir(src):                       # processor and template files the server expects; never the base shard index
    if f.endswith((".json", ".jinja", ".txt")) and not f.endswith(".index.json") and not os.path.exists(os.path.join(out, f)):
        shutil.copy(os.path.join(src, f), os.path.join(out, f))
print("MERGED MODEL SAVED:", out, "| files:", sorted(os.listdir(out)))