zhangxichen

Results 6 issues of zhangxichen

Hello and thanks for publishing the code of your research. while running experiments, I had some questions. In function get_input(), I notice that when SMILES length is longer than 218...

在api中,问答使用的是question,生成使用的是prompt,区别在哪里呢?

Version 1.0.11 fixed log in issue, but it still cannot sign in new account to get a local auth.

![image](https://user-images.githubusercontent.com/58805028/237045494-2176097b-6ef9-44fc-a54b-1721a2981b0d.png) docker配置完毕后默认账号密码登录不上去。。 求解 多谢

感谢您的工作~ 我观察到对于clip来讲,计算loss的时候是将对角线设置为1,其余全是0,;但是对于当样本类别少于batchsize的时候,即一个batch中会出现相同的文本标签,这个时候需不需要相应的将相同文本标签对不同图片的目标值设置为1?这似乎变成了一个多标签的优化问题,需不需要将loss函数变为BCELOSS? 我在这样操作的过程中在我的数据集上进行微调后,相似度会出现负值,这是合理的么?

Thanks for your amazing work~ For some reason, I want to try larger model (owlvit-large model) or latest model (owlv2) to fit my tasks, however when I modified function "load_model"...