import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel, LoraConfig
import os
from huggingface_hub import HfApi, ModelCard

# Load base model and your fine-tuned LoRA adapters
base_model = "mistralai/Mistral-7B-v0.1"
adapter_path = "./lora-adapters"  # Your trained LoRA adapters

model = AutoModelForCausalLM.from_pretrained(
    base_model,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Load adapters
model = PeftModel.from_pretrained(model, adapter_path)

# Save adapters for upload
model.save_pretrained("./hf-upload", safe_serialization=True)

# Save tokenizer
tokenizer = AutoTokenizer.from_pretrained(base_model)
tokenizer.save_pretrained("./hf-upload")