The training mode does not include 'masked-diffuse-lm'

[Orange1999]

Thanks for this creative work.

But I have encountered some problems in use, and I hope to get your help.

When I use the following training command as you provided :

python scripts/run_train.py --diff_steps 500 --model_arch bert --lr 0.0003 --lr_anneal_steps 400000 --seed 0 --noise_schedule sqrt --in_channel 128 --modality roc --submit no --padding_mode pad --app "--predict_xstart True --training_mode masked-diffuse-lm --roc_train ../datasets/ROCstory " --bsz 64

the losses is None

And I found that the parameter training_mode="masked-diffuse-lm" is not allowed in ./improved-diffusion/improved_diffusion/gaussian_diffusion.py.

This project code seems to be the same as Diffusion-LM, can you check the latest commit?

[Ma-Weijian]

met the same issue. anyone can tell me how to fix that?

[vikyou]

met the same issue. anyone can tell me how to fix that?Thank you very much！

[daniellaye]

Same question. Could I please ask where the masked-diffuse-lm loss is? Thanks a lot!

[vikyou]

Has anyone solved this problem?

[vikyou]

Does anyone email the author about the code？