Peter Hu

Results 6 comments of Peter Hu

hello hakaboom!非常感谢❤您的pr,我们对您提供的代码进行了基准测试,测试的结果如下。 1 以下代码段在部分情况会报错 https://github.com/hakaboom/Airtest/blob/d5d632ebfbb5b6131505b3cf763ee6acb8ced8e7/airtest/aircv/keypoint_base.py#L204 https://github.com/hakaboom/Airtest/blob/d5d632ebfbb5b6131505b3cf763ee6acb8ced8e7/airtest/aircv/keypoint_base.py#L214 2 在我们的测试数据中表现,相比当前sfit算法 漏检减少、误检增多,可能导致用户现有脚本的表现异常。 测试样本量总数 8372,以下为测试结果 正确 4260 + 2719 检测错位 344 漏检 801 虚警 247 正确 4187 + 2535 检测错位 634 漏检 694 虚警...

这些是测试的原始图像 方便您进行问题排查 ![screen_560005_tpl1586942359447](https://user-images.githubusercontent.com/10624004/173018178-6f507f74-d968-4ae4-8423-a49e2787ce9f.jpg) ![screen_560311_tpl1481082239753](https://user-images.githubusercontent.com/10624004/173018139-853a1a7e-ad04-4d38-94d7-f2caf5c8180c.jpg) ![screen_560402_tpl1464535754810](https://user-images.githubusercontent.com/10624004/173018145-43a0c2bb-ee2a-4c01-afef-6b1651ac038d.jpg) ![screen_561527_tpl1598925325189](https://user-images.githubusercontent.com/10624004/173018147-26715397-f1e8-4d37-b6eb-34e64c5b8628.jpg) ![tmpl_560005_tpl1586942359447](https://user-images.githubusercontent.com/10624004/173018154-056272ad-8313-4f98-9b84-325c08568fc3.jpg) ![tmpl_560311_tpl1481082239753](https://user-images.githubusercontent.com/10624004/173018163-1975f721-91cf-4493-93d4-a17563dccbfd.jpg) ![tmpl_560402_tpl1464535754810](https://user-images.githubusercontent.com/10624004/173018170-346087eb-331f-41cc-a576-055cb336c06f.jpg) ![tmpl_561527_tpl1598925325189](https://user-images.githubusercontent.com/10624004/173018172-6df1ca82-92f4-4c15-be10-fc2963cb3e25.jpg)

I guess the most problem is that you cannot install atari-py . Maybe you can try this https://github.com/j8lp/atari-py . it seems work for [his win10](https://github.com/openai/atari-py/issues/5#issuecomment-307647127)

Here is my change: https://github.com/Da-Capo/seed_rl/commit/a05ca9bd2daf26bccb020d53f356833b48c78783 If I use `temp_grads`, it will get an valueerror, but the `clip_grads` works well, so I wonder is the `clip_grads` synchronized correctly. ``` def apply_gradients(_):...

这个数据集以前存在实验室的电脑里,后来没再做相关的东西就没有备份,现在已经找不到了,实在不好意思。

这个代码也算是上古黑历史了[捂脸],不排除版本问题。我查了下这个报错是生成了空的词袋,你可以试着打印下 cut_docs 的内容确定是不是分词出了问题,或者看下CountVectorizer()的参数是不是有问题,还不行的话再向上排查。