Chao Li

I noticed that there are two implementations of ArcFace head. One is in head/metric.py, and the other one is in backup/metric.py. It seems that the former one split the weight...

In train_patch_clf_im4096_256_3Cls.sh , there is one param for train_patch_clf.py: --bias-multiplier 0.1 . But in patch_clf_train.py, there is not such a parameter. So are the train_image_clf.sh and the image_clf_train.py. What does...

Wrong returned images when using the Android app

2

Hello, I have downloaded and installed the Android app in my phone. But when I tried some text prompt, the returned image was always the first one. I am wondering...

梯度累积中的问题

1

关于梯度累积的代码，有几点不太明白，想请教一下。 1. accum_image_features 和accum_text_features已经得到了，为什么在get_loss中又要重新计算一个batch的特征。这样的话，所有的特征都被计算了两次，造成资源的浪费 2. get_loss进行了accum_freq次，但是每次的accum_image_features和accum_text_features并没有发生变化。loss的计算是否是重复了？ 相关代码如下： # First, cache the features without any gradient tracking. with torch.no_grad(): with autocast(enabled=(args.precision == "amp")): chunk_image_features, chunk_text_features, _ = model(images, texts) accum_image_features.append(chunk_image_features) accum_text_features.append(chunk_text_features)...

Why the mxnet2caffe.py exit when the layer is "running_mean"?

3

Hi, I have run the json2prototxt.py successfully, but when I run the mxnet2caffe.py, it occurs something wrong. The program exits when it meets the layer "mobilenet0_batchnorm0_running_mean". I don't know why...

Hello, thank you for your great work. But I run the face parsing model on the CelebAMask-HQ test set (2824 images) and got the below performance. The model I run...