Closed youngstu closed 1 year ago
Sorry I am not sure about the reason for this... Did you use the same environment as what we mentioned?
All have the same environmental configuration. Would it be convenient for you to pull and test the code again. Other friends have also encountered similar problems. Maybe there was a small problem with the code.
Very thanks.
https://github.com/Doubiiu/CodeTalker/issues/10 It look like the same problem.
OK I will pull and test the code again. Did you use the provided dataset preprocessing script? It should be the same as that in FaceFormer.
Very Thanks.
Yes, I am using the preprocessing script provided from CodeTalker.
Faceformer training is correct.
Found the cause of the problem, it's data_verts.npy bug. There is an issue with the data version of verts.npy(md5: 01bbeb78742afd15efe88ca318c7932d ). Early version training may have NAN issues, but the new data_verts.npy (md5: 2242f58f4697b046db7b3fd0cc7add59) version is right.
Very thanks.
OK I will close this issue then
sh scripts/train.sh CodeTalker_s1 config/vocaset/stage1.yaml vocaset s1 sh scripts/train.sh CodeTalker_s2 config/vocaset/stage2.yaml vocaset s2
On the vocaset training, Nan appears in both the first and second stages.