Closed kuldeepbrd1 closed 1 year ago
Thanks @kuldeepbrd1 .
You are right! This would be highly benefitial to sample properly the grasps. If you feel up to it, create a pull request, I tested it and if everything works smoothly, I accept it.
Thanks alot :)
The scripts in
scripts/sample
do not use n_grasps from the cli arguments.n_grasps
andn_envs
in not used. Instead it always generate number of grasps equal to batch size specified inget_approximated_grasp_diffusion_field(...)
, as here: https://github.com/TheCamusean/grasp_diffusion/blob/3a2cb1448270798435479ee7cf8d1fbd9d5127c5/scripts/sample/generate_partial_pointcloud_6d_grasp_poses.py#L28Batchwise sampling of
n_grasps
would be nice. This also avoids CUDA/cpu memory errors when largen_grasps
is high. Something like this inmain(...)
:and
get_approximated_grasp_diffusion_field(...)
changed toIt's not super critical to add this to code, so I highlight here. (Also, happy to also create a pull request, if you require)