pytorch Version

[jiaszhang]

Hello, I need to compile pytorch, could you tell me the pytorch version the project ued? thank you!

[chaoyuaw]

Hi, Sorry for the late reply. I don't remember the exact version I used when I developed this repository (April 2019), but I used the latest version back then, so my best guest would be around v1.0. Please let me know if you have any issue using this repo. Thanks!

[luuckiest]

I have problem installing the newest version of pytorch.
Ubuntu is 18.04,
Cuda=9.0
Cudnn = 7.5
gcc = 5.5
python =2.7.x

If you could let me know which version are you using?

[chaoyuaw]

Hi, I'm sorry but I don't have access to the environment I used to develop this repo, so I don't have answers to your questions. However, if it's a PyTorch installation question (as opposed to a LFB-specific question), maybe "https://github.com/pytorch/pytorch/issues" would be helpful.

[Holmes-GU]

@chaoyuaw Hi, I am reading your codes and I am not sure about the scale operation, in the code, there is theta_phi = model.Scale(theta_phi, theta_phi, scale=latent_dim**-.5). What dose it mean? Is every element divided by scale? Thank you.

[chaoyuaw]

Yes, it's element-wise scaling by a factor of (1 / sqrt(latent_dim)). We follow Sec-3.2.1 & footnote 4 of https://arxiv.org/pdf/1706.03762.pdf for this design choice.

[Holmes-GU]

Thank you very much. | | SJTU 顾智浩 邮箱：[email protected] | 签名由 网易邮箱大师 定制 On 07/15/2020 12:25, Chao-Yuan Wu wrote: Yes, it's element-wise scaling by a factor of (1 / sqrt(latent_dim)). We follow Sec-3.2.1 & footnote 4 of https://arxiv.org/pdf/1706.03762.pdf for this design choice. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.