YE Zhou

The feature files is too large to upload and you can follow [this instruction](https://github.com/ruotianluo/self-critical.pytorch/blob/master/data/README.md#convert-from-peteanderson80s-original-file) to prepare the bottom-up features.

I git clone the code and it can run correctly. You can set a breakpoint in `/users/vlad/image_grounding/Grounded-Image-Captioning/misc/rewards.py", line 68` and check why you got the `key 29` (instead of 28)...

Ok. Thanks!

import nltk nltk.download('punkt')

Thanks!

Thank you! I found the deploy.prototxt file. But I still have some confusion as follow: If you extract features at the layer named by "fc-conv7" (or "relu7") and the size...

Thank you very much!

Hi @maryawwm , my env is the linux os, it seems that you are using the windows version. Anyway, this RuntimeError is caused by the unexpected dtype of torch.embedding input,...

Hi @maryawwm , first you can check whether you are using pytorch 1.6. Second, if you have multiple GPUs on your machine, you can temporarily set `CUDA_VISIBLE_DEVICES=0` and try again.

You can also refer to [Multi-GPU Error](https://github.com/pytorch/pytorch/issues/40457)