shallweiwei
shallweiwei
Of course.But I don't see your answer on github ---Original--- From: ***@***.***> Date: Mon, Dec 12, 2022 19:04 PM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [saic-mdal/lama] how to use the...
You can try to use absolute path.
Hi,everyone. It seems this [link ](https://hkustconnect-my.sharepoint.com/:f:/g/personal/xliea_connect_ust_hk/EsV9Q8s2g3ZNgwSYEfdnvTAB4_uIq8LEwXr1VcvaB0X8bw?e=cKtw5j)can work. I don't know what happened, but I also encountered above problems at first. I guess it's because the link provided by @fh2019ustc is...
I can share a new way here.You can use detach when calculate dfake for D_loss().And after D_loss had backward,then calculate another dfake using discriminator for G_loss().
@htzheng Hello, could you release the training and inference code ?
Both on v1.0 and v1.1
 I used Scribble Model. I tried to draw mountains,but the output is really bad. (I also tested Canny Model.It seems 256 resolution is not a good choice)
I believe it's this line `model.load_state_dict(load_state_dict('./models/control_sd15_canny.pth', location='cuda'))` Does anyone have the same problem?Sometimes it's very fast```
I use my own dataset, but mine is in a variety of colors, mostly a bit of light green
> > I use my dataset to train and notice that the masks usually have different colors than the parts around them.They are easy to distinguish with naked eyes. So...