MagicSource

thanks for your reply. I need Chinese pretrained model so I'd better stick with tensorflow. I tried tf1.14 it outputs some error but I have fixed it. And I have...

Thanks for this helpful suggest, it's useful, but I got an error when start service: ``` ./bertsvr.sh 2 master!? start BERT server... 2020-04-06 12:52:18.136044: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library...

Fxied it, when test got an error: ``` I0406 13:02:47.061372 140011026507520 __init__.py:537] ready and listening! ^TI:PROXY:[htt:enc: 47]:new request from 127.0.0.1 I0406 13:04:28.258268 140012192503552 http.py:47] new request from 127.0.0.1 None E:PROXY:[htt:enc:...

Fixed it, the model whatever inputs are, output label always 0...    this is eval result on my model: ``` eval_accuracy = 0.9783315 eval_f1 = 0.97806644 eval_loss =...

No, I am using default corpus which is weibo_sentiment100k to train. You can test my sentence, the model can not handle this situations.

thanks!

@Nor-s hi, thank u for the reply, do u have an example for this? How should I get the R and S from a SMPL model and a Mixamo model...

@Nor-s thank, does there a way make your code buildable on macOS? I wnanna have a try, this seems exactly what I need

@Nor-s hello, I am exactly Sillcon version M1 mac. But you needn't build arm python yourself. Using mini-forge can have a arm python in conda. Have you tride that? this...

however, metahuman face using 51 blendshape as apple's ARKit.