kas-one

@fengyang0317 Thanks for your quick reply, I used the model-10000 which you provided and obtained the results in the paper. Is this model using obejct2sent initialization? I tried the model-20000...

yes, I just download the code without polishing it. But I use batch_size 256 without multi-gpu, is this important?

Thanks, I will try 5e-4. ------------------ 原始邮件 ------------------ 发件人: "Yang Feng"; 发送时间: 2019年6月28日(星期五) 中午11:10 收件人: "fengyang0317/unsupervised_captioning"; 抄送: "旋风小子";"Author"; 主题: Re: [fengyang0317/unsupervised_captioning] about the reproduce ofmetrics reported in paper? (#4) I...

and, I found a small problem. The size of the my trained model/graph file is a little different from yours as below.  yours  ours(I directly run the command....

I use tf 1.12， but without problems of training and evaluation.

@fengyang0317 Thanks for your patient help, I solve this problem successfully.

@fengyang0317 When I trained for second time , everything is OK. My analysis is as below, maybe it's wrong. I found that the test metrics are very related with the...

@songpipi I means that I delete the models and rerun the code again without any changes. I advice that you can test more models, i.e. 20000,21000,22000...., only one may have...

@songpipi since only one model may deviate from stable value. And another advice is to delete the models and rerun the image_cap_full.py again.

@songpipi the stable value = policy rewards. The name in the tensorboard is loss, and the corresponding name in the file the author provided is loss1.