DRSY EMO issues - Githubissues

DRSY / EMO

[ICLR 2024]EMO: Earth Mover Distance Optimization for Auto-Regressive Language Modeling(https://arxiv.org/abs/2310.04691)

111 stars 13 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

怎么用于普通分类任务？

#13 justajustin closed 1 month ago
0
是否会导致模式坍塌？

#12 980202006 closed 5 months ago
4
Is there a good way to initialize cost matrix when pretraining from-scratch?

#11 DaehanKim closed 5 months ago
2
Fix typo in README

#10 tomaarsen opened 6 months ago
0
Is it possible for you to cite "Learning with a Wasserstein Loss" in the camera-ready version?

#9 YouJiacheng closed 7 months ago
1
emo_loss 会变成nan

#8 zzf-damon opened 7 months ago
5
关于emo loss的疑问

#7 Vincent-Poirot opened 9 months ago
1
gt_q = (q_grad * one_hot).detach()

#6 chenxu001 closed 9 months ago
3
代码中提供了三种DEMD和MLE的方式, 请问论文中使用的是哪一种呐?

#5 oyjxer closed 9 months ago
1
分布式多机训练, loss 训练 300 step 后会变成负数

#4 jiaruipeng1994 opened 9 months ago
3
Add return_dict option to be compatible to llama

#3 xufangzhi closed 10 months ago
0
update README.md

#2 eltociear closed 10 months ago
0
请问训练的时候loss上升是正常现象吗

#1 lichen914 closed 10 months ago
7