wsh2836741

Results 7 issues of wsh2836741

Hi,thank you for your contributions.I want to use your faster rcnn,can you please tell me where are the caffemodels for train and test?

感谢您非常出色的工作。有个小问题想问一下,看过代码后发现loss损失函数的实现和nt-xent损失函数的公式不太一致?

你好,麻烦问一下shuffle操作是不是会有问题,比如”蛋糕奶油“和”奶油蛋糕“,经过这种shuffle后是不是体现不出位置重要性了。

Thanks for your good job,and does the model support the task in Chinese?

感觉团队杰出的工作。关于Task-specific Distillation第一步中间层蒸馏,比如我是分类任务,由于中间层蒸馏不会训练最后的分类层参数,所以想问下第一步中间层蒸馏的评价指标是什么?还是说不需要关注评价指标,只看loss下降,模型收敛就可以?非常感谢!

Hi,thanks for excellent job,can it support Chinese entity linking(both in model and database)?

Thanks for excellent work.But i am confused,it said 'out of memory' when i use titan xp GPU(12G).And the largest scale i can use is 1100*1100.Thank you in advice.