As you have already done the training, you can now load the model directly without using the LoraConfig, as follows:

# Load model directly

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("mustangs007/gemma-medical_qa-Finetune")

model = AutoModelForCausalLM.from_pretrained("mustangs007/gemma-medical_qa-Finetune")