Self-critical sequence training

[Stephanie-Shen324]

Hi,

First of all, thank you for sharing your excellent work and nicely-written code!

I'm wondering if you have tried implementing self-critical sequence training for this model? I am currently trying to implement it on Google Colab. But I always get a CUDA Out Of Memory issue. If you have tried, did you encounter similar issues?

Best wishes,
Stephanie Shen

[zhjohnchan]

Dear @Stephanie-Shen324,

MAKE SURE you set torch.with_no_grad() when you sample a baseline report when performing SCST. This always causes the CUDA OOM problem.

Best,
Zhihong

[wlufy]

Hi,

First of all, thank you for sharing your excellent work and nicely-written code!

I'm wondering if you have tried implementing self-critical sequence training for this model? I am currently trying to implement it on Google Colab. But I always get a CUDA Out Of Memory issue. If you have tried, did you encounter similar issues?

Best wishes, Stephanie Shen

Hi, sorry to disturb you.
I am trying to implement self-critical sequence training for this model. But i have some difficulties, i can't understand the differences between labels and gts. Like the following:
def forward(self, fc_feats, att_feats, labels, masks, att_masks, gts, gt_indices, sc_flag, struc_flag, drop_worst_flag):
Could you share the code with me? Thanks!