Wicknight

@Little-wen 您好！感谢您对RecBole-CDR的关注！RecBole-CDR的评测框架继承自RecBole，目前仅支持对目标域做评测，如果您想要同时对源域和目标域做评测，您需要重写Trainer中的evaluate函数，同时保存两个数据域的data_struct，分别进行评测。

Hello @hv-abacus , It seems that you are not filtering the data. The dataset statistics that you report [here](https://github.com/RUCAIBox/RecBole-CDR/blob/main/results/Amazon.md) were obtained after 10-core filtering, which were specified by parameters '**user_inter_num_interval**'...

您好 @zzhuncle ，RecBole-CDR是支持加入自己编写的模型的，您需要在'RecBole-CDR/recbole_cdr/model/cross_domain_recommender'下添加自己的模型，RecBole-CDR的框架结构是沿用自RecBole，您可以查看RecBole的使用文档：[v1.0.1版本](https://recbole.io/docs/v1.0.1/)(RecBole-CDR使用的版本)和 [v1.1.1版本](https://recbole.io/docs/#)。也欢迎您通过pull requests将您实现的模型整合到我们的框架当中。

Hi @ajaykv1 Since the training process involves multiple datasets (source datasets, target datasets), we do not support specifying with just one parameter of the dataset. In contrast, we set the...

Hello @kavita-rk Could you please give me more detailed information? It seems that I can't tell from your description why the error occurred.

@AML-CityU 您好，感谢您对RecBole-CDR的关注！ 需要向您确认几点信息： 1. 您在测试单域ctr模型时是否使用了额外的特征？RecBole-CDR中的模型并未使用任何内容信息； 2. 您在测试单域ctr模型时是否也测试了一些top_n推荐模型的效果？因为这些跨域模型的目标实际都是做top-n推荐的，对特征的处理不如ctr模型高效（如deepfm），推荐您将这些模型和BPR等模型进行一次对比，这样可能才是一次公平的比较。

@yoosan Hello, thanks for your attention to RecBole-CDR! Actually we only provide support for `full` mode in current release. If you want to use `uni999` evaluation, you need to write...