Zhuowei Chen

Here is the code I used [darts-rnn](https://github.com/fortunechen/mydarts-rnn) with pytorch1.7+cuda11.0

I will try reproduce the result on `pytorch 0.3.1` again.

> Empirically, darts search time and performance are often inversely proportional and the early stop is usually inevitable. I agree so. However the total search time is the same as...

> I will try reproduce the result on `pytorch 0.3.1` again. Here is the result without any modification of original code. 

> I also improved the code to make it compatible with PyTorch 1.1 while allowing multi-GPU training on both RNN and CNN experiments.~ you can refer: https://github.com/alphadl/darts.pytorch1.1 hi, the weight...

The above only contains image feature. The below is for coco-caption experiments, It can be used for reproducing the results reported in the paper. containing experiment configs, training/val/test image features,...

> hi, Do you know how the author generated the data provided in the link below(https://biglmdiag.blob.core.windows.net/vinvl/datasets/coco_caption)? I don't know, maybe you should open a new issue

same issue here, + 2line and downgrade requests not work

> https://github.com/switchablenorms/CelebAMask-HQ/tree/master/face_parsing#well-trained-model This label is not suitable for this model

> Made the necessary changes to port the code to python 3.6+ and pytorch 1.0+. > > https://github.com/davidstap/AttnGAN Thx, I can reproduce the results in this paper by your repo.