NameError: name 'round_up' is not defined[BUG/Help] <title>

[cuizhiyan]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

p-tuning method is used to tune the ADGEN dataset. The default super parameters are used.
so how to fix it?

delete " –quantization_bit 4 " works for me.
but i need to use it with quantization.

Expected Behavior

No response

Steps To Reproduce

bash train.sh

Environment

- OS:Linux
- Python:3.8
- Transformers:4.30.2
- PyTorch:2.0.0
- CUDA Support (`python -c "import torch; print(torch.cuda.is_available())"`) :True

Anything else?

No response

[Dicer-Zz]

Quantization rely on cpm_kernels. Install it by:

pip install cpm_kernels

[cuizhiyan]

pip install cpm_kernels

it works. Thanks a lot

[todochenxi]

Thanks!

[RonaldoPeng]

Quantization rely on cpm_kernels. Install it by:

pip install cpm_kernels

It really works. Thanks.