jacqueline-weng

For pretrained, I think you can just load each partial fc weight file separately for each gpu when training starts. The problem only occurs when you use different number of...

> i have same problem on pytorch 1.12.0 https://github.com/pytorch/pytorch/issues/42993 This issue implies that it has fixed in Pytorch1.11.0 but I haven't tried.

'logits' are the probabilities the current image belonging to different classes (one identity is a class). You may check ArcFace paper for details. https://arxiv.org/abs/1801.07698

> So you mean that the output of the model should be a softmax and not an embedding ? i fixed the output of my ViT to 512 thinking that...

``` import torch from losses import ArcFace batch_size = 2 num_classes = 10 logits = torch.randn((batch_size, num_classes)) labels = torch.tensor([[5],[6]],dtype=torch.long) ArcFaceLoss = ArcFace() loss = ArcFaceLoss(logits, labels) print(loss) ``` 1....

First of all, answering questions helps me better understand the whole stuff, I'm willing to share and discuss to learn more. I think you should spend time reading the code...

The performance of VIT and R50 is affected by many variables. 1. First, to make both training converge, you need to have a proper, often large batch size, which may...

I also applied vit_t and had loss stuck at 22 for quit a few epochs.   I stopped the training and restarted it with SGD to see any change.

> Hi, [jacqueline-weng](https://github.com/jacqueline-weng) , have you tried this？ > > https://raw.githubusercontent.com/anxiangsir/insightface_arcface_log/master/wf42m_pfc02_40epoch_8gpu_vit_t/training.log > > https://github.com/deepinsight/insightface/blob/master/recognition/arcface_torch/configs/wf42m_pfc03_40epoch_8gpu_vit_t.py Thank you for replying me and showing me the vit_t training result. Things were a bit...

Thanks @anxiangsir, I'm trying with accumulative gradient. An error occurs at the first backward saying some parameters in the module are marked ready to reduce twice. After changing the module...