free_wind
free_wind
I have the same issue, is there any update? It seems the pretrained CLIP model fails to provide meaningful results.
Thanks for pointing out this issue. In our released code, we found that positional encoding affects the final performance insignificantly (about 0.5%) but would lead to instability during training and...
您好，感谢您对我们工作的支持。这一部分模块可能会占用比较大的CPU线程消耗，在CPU计算量不足的情况下，每次迭代的part搜索会导致速度变慢。由于当时使用服务器进行计算，并未发现明显速度影响。您可以尝试调整部件个数来极大缩短计算时间。：）
Hi, I have tested it on several different machines and found it works well. Besides the commonly occurring Nan training in other codes, one possible problem here might be the...
Hi, modifying the hyper-parameters of part discovery (part number and queue number) may help.
cluster_id.txt不需要使用，抱歉
> 很抱歉再次打扰您，不知道为什么我的数据加载是成功的，但是几乎学习不到任何的特征 loss是否正常下降，现象是什么呢。多训练一段时间观测一下准确率？没遇到类似情况，已经换了多台机器测试