Closed HamadYA closed 1 year ago
Hi, Do you have the output of knowledge distillation 'npz' or 'tfrecord' file in a link, please?
$ CUDA_VISIBLE_DEVICES='0' ./data_distiller.py -M subcenter-arcface-logs/r100-arcface-msfdrop75/model,0 -D /datasets/faces_casia_112x112_folders/ -b 32 --use_fp16 Output: faces_casia_112x112_folders_shuffle_label_embs_normed_512.npz
Sorry but I'm not keeping that either. How about using the ported h5 models instead of the MXNet one?
h5
Noted will do Thanks!
Hi, Do you have the output of knowledge distillation 'npz' or 'tfrecord' file in a link, please?
$ CUDA_VISIBLE_DEVICES='0' ./data_distiller.py -M subcenter-arcface-logs/r100-arcface-msfdrop75/model,0 -D /datasets/faces_casia_112x112_folders/ -b 32 --use_fp16 Output: faces_casia_112x112_folders_shuffle_label_embs_normed_512.npz