wanlipeng

数据集下载

3

你好,请问下数据想离线下载使用,代码中提供的网址是谷歌的需要翻才能打开下载,有没有提供百度网盘的下载方式或者其他离线下载的途径呢?

Msra训练NER,报cuda 内存满了？

2

你好，我在用Msra数据训练NER任务时，batchsize＝8时训练就报cuda内存错误，batchsize＝4可正常训练，就是训练迭代次数特别多，使用四GPU2080显卡也是这样问题，分布训练也太慢了，想问下，Msra数据这么吃内存吗，有更好的办法解决没，求教？

NameError: name 'fused_act_ext' is not defined

4

尝试跑inference_stylegan2.py，出现"NameError: name 'fused_act_ext' is not defined"这个错误，有大佬遇到过，我看报错行，源码里面有used_bias_act.cpp和fused_bias_act_kernel.cu文件的

你好，用该跨框架训练中文场景OCR数据，训练SRN loss一直不会降低，更换ctcloss也是不行，大佬有训练过没，求指导交流交流？训练PAD论文是使用EOS填充，做交叉熵loss的时候ignor_index=PAD,字符行最后字符EOS求loss也忽略吗？

Invalid instace which ends at line:4093 has been drpoped

4

你好,我在训练开源数据weibo_NER数据时候可以这个正常训练,仿照weibo_NER数据格式做自己的训练数据训练报错,"Invalid instace which ends at line:4093 has been drpoped",加载数据时候出错,都是用"bio"encoding type,求指导下!

TypeError: make_grid() got an unexpected keyword argument 'range'

6

TypeError: make_grid() got an unexpected keyword argument 'range'

How to run a demo test quickly with a video

1

How to quickly run video sequence of an actual scenario. Could you please provide a demo test code?

dataset about "new optical flow data featuring photorealistic images and precise optical flow ground truth"

1

Please when the new optical flow data mentioned below about featuring photorealistic images will be released.

Abount CC3M Pseudo Annotations use

5

Hello, I have downloaded "CC3M-246K" annotations data, if I have downloaded CC3M image data, How can I match the image data and annotation data files? There is no URL address...

When using yolo-world training, I observed that the loss first decreased, then increased, and finally became nan. What might be the cause? Has anyone else encountered this situation? The log...