AutoTrain advanced CLI: error: unrecognized arguments: --fp16 --use-int4

Question

I am currently facing an issue while fine-tuning my data with the LLM model in a Colab notebook using the provided autotrain tool. The error message suggests that the arguments '--fp16' and '--use-int4' are not recognized by autotrain. I've checked the documentation and syntax, but the problem persists. Could you provide guidance on resolving this issue or offer insights into any potential solutions? Thank you.

/usr/local/lib/python3.10/dist-packages/torchvision/io/image.py:13:
 UserWarning: Failed to load image Python extension: '/usr/local/lib/python3.10/dist-packages/torchvision/image.so: undefined symbol: _ZN3c104cuda9SetDeviceEi'If you don't plan on using image functionality from`torchvision.io`, you can ignore this warning. Otherwise, there might be something wrong with your environment. Did you have `libjpeg`or`libpng`installed before building`torchvision` from source?   warn( usage: autotrain <command> [<args>] AutoTrain advanced CLI: error: unrecognized arguments: --fp16 --use-int4

screenshot of the error

this code was working fine till yesterday on the colab notebook given in this https://github.com/huggingface/autotrain-advanced repo for fine tuning LLM, Now getting this error.

Answer 1

use this quantization int4 and --mixed-precision fp16 this will solve your problem

Answer 2

Clone the Google Collab notebook again and execute as there was a code change that caused the old way of passing parameters to break.