Text Generation
Transformers
PyTorch
English
llama
Inference Endpoints
text-generation-inference

getting strange tokens after finetuning on Qlora

#11
by monuminu - opened

Hi All,

prompt = """### Instruction: write a python code to read csv

Response:"""

I am getting strange tokens like below

:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

Together org

Hi @monuminu , what hyperparameters did you use for training and generation?

I think the issue is because of monkey patch which was not applied.

Sign up or log in to comment