abcsimple

adaface issues

1

您好，在adaface的repo中看到您发布过一些issue，有相关问题想请教一下，请问能方便提供一下邮箱或微信账号吗？ 抱歉占楼，感谢

no such file or directory: /usr/src/app/.cache/huggingface/hub/models--laion--CLIP-ViT-H-14-laion2B-s32B-b79K/refs/main

6

I download open_clip_pytorch_model.bin from huggingface and saved it in ./laion/CLIP-ViT-H-14-laion2B-s32B-b79K, but got the erro: no such file or directory: /usr/src/app/.cache/huggingface/hub/models--laion--CLIP-ViT-H-14-laion2B-s32B-b79K/refs/main. I found that in file factory.py of open_clip lib, the...

Hi, dost partial fc v2 has any improvement on "training speed" or "GPU memory cost"? THX

Since the feature injection ratio is determined by the life indicator T_{y_i} and T_{y_i} is refreshed based on delta t, the feature injection ratio should be adjusted during the training...

hi, did you normolize the training data by mean and std values when you training the pretrained models?

Hi, I would like to express my admiration for your work. I am curious if you used model parallelization in SphereFace2, since the classification layers were not seperated to different...

Speed of training proceeds

1

Hi, I evaluated the speed in training proceeds between VanillaNet and ResNet, vanillanet has a little bit less params and flops than resnet. However, vanillanet is slower than resnet and...

Questions about hyperparameter λ

1

Hi! Your work is fantastic! I have two questions regarding the hyperparameter λ: According to Formula 1, λ should start at 0 and end at 1 during the training process....

Remove global average pooling out of "forward_head"

1

Great job on your work! I have a small suggestion regarding the code in inceptionnext.py. It would be more convenient if the line "x = x.mean((2, 3)) # global average...