Issue merging LoRa adapters back into gptq quantized model

[RonanKMcGovern]

System Info

A PEFT model cannot be merged with the quantized model and pushed to hub...The error is:

Cannot merge LORA layers when the model is gptq quantized

after trying:

model_id = "ybelkada/llama-7b-GPTQ-test"
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

from peft import PeftModel

# load PEFT model with new adapters
model = PeftModel.from_pretrained(
    model,
    adapter_model_name,
)

model = model.merge_and_unload() # merge adapters with the base model.

Note that 'model' is the gptq quantized model, origin
I had originally posted in autogptq but realise this may be a question for here.

### Who can help?

@fxmarty perhaps you could assist? Maybe there is an alternate approach

### Information

- [ ] The official example scripts
- [X] My own modified scripts

### Tasks

- [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...)
- [ ] My own task or dataset (give details below)

### Reproduction (minimal, reproducible, runnable)

See above

### Expected behavior

Typically, this approach of merging (like when I do it with NF4 quantization) will work.

[SunMarc]

Hi @RonanKMcGovern, this is not a bug. We still haven't implemented the merge_and_unload() for gptq quantized model. Feel free to request this as a feature enhancement in peft library.