kangpeilun / VastGaussian

This is an unofficial Implementation
Apache License 2.0
360 stars 26 forks source link

training failed #33

Open yuxiaoz opened 2 months ago

yuxiaoz commented 2 months ago

Hello, I train the building dataset on my native GPU RTX3060-6GB, according to the log, the 9(3x3) partitions have been trained, but no "point_cloud" folder is saved in the work space. Then, the seamless_merge processing is skipped. Is the GPU performance too low thus causing the training crashed?

picture