Unpickling stack underflow when creating FLAME decoder

[jasonlee328]

Hi, amazing work on PointAvatar. I'm getting the issue where there is a pickle.unpicklingerror when training the data. I ran IMavatar's training process multiple times without getting the error. Is there a way to fix this?

shell command : scripts/exp_runner.py --conf ./confs/subject1.conf Loading data ... Loading all images, this might take a while. Loading all images, this might take a while. Finish loading data ... creating the FLAME Decoder 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1111/1111 [00:21<00:00, 52.12it/s] 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1639/1639 [00:32<00:00, 50.49it/s] Traceback (most recent call last): File "/home/jaslee20/PointAvatar/code/scripts/exp_runner.py", line 27, in runner = TrainRunner(conf=opt.conf, File "/home/jaslee20/PointAvatar/code/../code/scripts/train.py", line 101, in init self.model = PointAvatar(conf=self.conf.get_config('model'), File "/home/jaslee20/PointAvatar/code/../code/model/point_avatar_model.py", line 29, in init self.FLAMEServer = FLAME('./flame/FLAME2020/generic_model.pkl', './flame/FLAME2020/landmark_embedding.npy', File "/home/jaslee20/PointAvatar/code/../code/flame/FLAME.py", line 48, in init ss = pickle.load(f, encoding='latin1') _pickle.UnpicklingError: unpickling stack underflow