Error while loading MISTRAL LLM for fine-tune. Qlora doesn't work but full works

if I try to load the model in this way :

bnb_config = BitsAndBytesConfig(    load_in_4bit=True,    bnb_4bit_quant_type="nf4",    bnb_4bit_use_double_quant=True)model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.2",    quantization_config=bnb_config,    torch_dtype=torch.bfloat16,    device_map="auto",    trust_remote_code=True,    token=access_token    )

it gives me this error: ValueError: Some modules are dispatched on the CPU or the disk. Make sure you have enough GPU RAM to fit the quantized model. If you want to dispatch the model on the CPU or the disk while keeping these modules in 32-bit, you need to set load_in_8bit_fp32_cpu_offload=True and pass a custom device_map to from_pretrained. Check https://huggingface.co/docs/transformers/main/en/main_classes/quantization#offload-between-cpu-and-gpu for more details.

If I don't use any quantization_config the model load without problems... but I get OOM error after while trying to fine-tune it.

I'm working on jupiter notebook with 100GB VRAM.

Do u have any idea ?

I tried other configs but it works only without further config.