Open xiaoqidaov2 opened 2 months ago
train_mouth这里报错是什么原因 tb_writer.add_images(config['name'] + "view{}_mouth/depth".format(viewpoint.image_name), (render_pkg["depth"] / render_pkg["depth"].max())[None], global_step=iteration)
/root/anaconda3/envs/TalkingGaussian/lib/python3.8/site-packages/tensorboardX/summary.py:286: RuntimeWarning: invalid value encountered in cast tensor = (tensor * 255.0).astype(np.uint8) NaN or Inf found in input tensor
debug后发现
render_pkg["depth"] tensor([[[0., 0., 0., ..., 0., 0., 0.], [0., 0., 0., ..., 0., 0., 0.], [0., 0., 0., ..., 0., 0., 0.], ..., [0., 0., 0., ..., 0., 0., 0.], [0., 0., 0., ..., 0., 0., 0.], [0., 0., 0., ..., 0., 0., 0.]]], device='cuda:0')
找到问题了，lms 存在大量负数
xiaoqidaov2
commented
2 months ago
train_mouth这里报错是什么原因 tb_writer.add_images(config['name'] + "view{}_mouth/depth".format(viewpoint.image_name), (render_pkg["depth"] / render_pkg["depth"].max())[None], global_step=iteration)
/root/anaconda3/envs/TalkingGaussian/lib/python3.8/site-packages/tensorboardX/summary.py:286: RuntimeWarning: invalid value encountered in cast tensor = (tensor * 255.0).astype(np.uint8) NaN or Inf found in input tensor
debug后发现
render_pkg["depth"] tensor([[[0., 0., 0., ..., 0., 0., 0.], [0., 0., 0., ..., 0., 0., 0.], [0., 0., 0., ..., 0., 0., 0.], ..., [0., 0., 0., ..., 0., 0., 0.], [0., 0., 0., ..., 0., 0., 0.], [0., 0., 0., ..., 0., 0., 0.]]], device='cuda:0')