Hi everyone, I am meeting this problem when I try to deploy Qlora LLAMA 2 13B on SageMaker and cannot find the solution for it. When I merge the adapter with the base model, there is no inv_freq, but when it comes time to deploy, somehow this layer requirement appears. I have tried to fix the bin file,… but none of those worked. Is this an error from HuggingFace or Sagemaker lib?
