error while training about unilm HOT 10 CLOSED

@DivyaWadehra

cd ~/code/unilm/src
pip install --user --editable .

from unilm.

@DivyaWadehra Copy the folder contained in src directory named 'pytorch_pretrained_bert' to 'biunilm'.

from unilm.

ModuleNotFoundError: No module named 'apex.optimizers'

and I am running the code on CPU

from unilm.

#23

from unilm.

I am ruuning code for question generation, that's for summarization
I am not using fp flag

from unilm.

The code change should be the same for both qg and summ in order to run the program on cpus.

from unilm.

By default it is taking fp into consideration how to disable the use of it while training QG model

from unilm.

In order to run the code on CPU, we can remove the --fp16 and --amp flags. Moreover, we need to delete .cuda(), and apex package usage.

from unilm.

But in the script itself we are not using these flags, can you generate a code for cpu?

from unilm.

Hi
I am getting this error, can you help?

ModuleNotFoundError: No module named 'fused_layer_norm_cuda'

from unilm.