Open muyinjiangxue opened 4 years ago
@bbrattoli dear Biagio, I am suffering the same issue. Would you please help to fix it?
Is it get_datasets() instead of get_dataloaders() @muyinjiangxue
Thanks a lot!
hello @nanguoshun I am suffering the same issue. datasets = dataset.get_datasets(opt) or dataloaders = dataset.get_datasets(opt)
can you tell me which one is correct ? i tried both but still get the error
Total batch size: 22
UCF101: total number of videos 13320, classes 101
HMDB51: total number of videos 6766, classes 51
Traceback (most recent call last):
File "main.py", line 65, in
hi @KP1-cmd , I have no idea for the issues. I suggest we send emails to authors for the issues.
Thanks a lot!
hi @KP1-cmd , I have no idea for the issues. I suggest we send emails to authors for the issues.
Thanks a lot!
Ok thank you
@muyinjiangxue did you fix the error ?
The issue comes from the fact that “get_kinetics()” in auxiliary/auxiliary_dataset.py, needs to be adapted to your format used for the Kinetics dataset. When I implemented this code I used a dataset from the AWS team and I am not allowed to publish that. I suggest you fix the get_kinetics and debug the dataloader accordingly. Unfortunately Kinetics is not an easy dataset to work with.
A quicker solution is to download one of the model that I trained and only evaluate or fine-tune on UCF101.
Thank you for pointing this out, I will mention it in the readme.
Cheers, Biagio
P.S. Sorry for the late reply
Your answer makes me understand. Thank you very much for your reply!Everything is very good
------------------ 原始邮件 ------------------ 发件人: "Biagio Brattoli"<notifications@github.com>; 发送时间: 2020年6月14日(星期天) 晚上9:51 收件人: "bbrattoli/ZeroShotVideoClassification"<ZeroShotVideoClassification@noreply.github.com>; 抄送: "西瓜她说爱夏天"<462001974@qq.com>;"Mention"<mention@noreply.github.com>; 主题: Re: [bbrattoli/ZeroShotVideoClassification] module 'dataset' has no attribute 'get_dataloaders' (#2)
The issue comes from the fact that “get_kinetics()” in auxiliary/auxiliary_dataset.py, needs to be adapted to your format used for the Kinetics dataset. When I implemented this code I used a dataset from the AWS team and I am not allowed to publish that. I suggest you fix the get_kinetics and debug the dataloader accordingly. Unfortunately Kinetics is not an easy dataset to work with.
A quicker solution is to download one of the model that I trained and only evaluate or fine-tune on UCF101.
Thank you for pointing this out, I will mention it in the readme.
Cheers, Biagio
P.S. Sorry for the late reply
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.
dataloaders = dataset.get_dataloaders(opt)
The data loading method here is not seen in dataset.