rcg
rcg copied to clipboard
LTH14
PyTorch implementation of RCG https://arxiv.org/abs/2312.03701
imagenet对我来说太大了，我想用自己的图片来进行训练，我是不是直接把八万张图片丢到train文件夹里面就可以了呢，请问一张a100够吗，我需要调整哪些别的超参数呢，谢谢您的工作，祝您一切顺利！！！
Hi Tianhong. There are always errors when training `main_rdm.py ` regardless of the value setting of batch size, e.g. , from 32 to 512. What' s the problem and how...
As there is no code related to figure7 interpolation in the GitHub repository.
Dear author, You have shown good results in generating data distribution like ImageNet, CIFAR-10 and iNaturalist, which are category-based datasets. I wonder whether you have experimented on generating more complex...
Impressive work, thank you for sharing! I have now reproduced most of the experiments; additionally, I plan to fine-tune on a small-scale dataset. Could you provide the moco-v3-large code? I...
Thx for impressive work and code. So, after checking the inference code under the case conditioned by an image, the whole generation just degenerates into a SSL representation conditioned MAGE....
@LTH14 I have not yet quite understood that why use `zero_module` in ResNet and SimpleMLP , it just zero out directly the tensors (input_rep_tensor + time_tensor+condition_tensor) ( torch.all(h==0)=> True), which...
I noticed that a new projector head MLP is added after loading the pre-trained MoCo v3 model. However, the parameters of this newly added component are also set to requires_grad=False....
Does anyone know how to generate the visual results in Figure 6? I see that they extract SSL representations from image samples, and the authors don't seem to describe how...
Thanks for your excellent work! I am facing difficulties while training RDM, and I hope to receive your assistance. When training RDM, in ddpm.py, at line 564 in the get_input...
