Closed im-renpei closed 8 months ago
To get numbers from the paper, run the benchmark script following: https://curobo.org/source/getting_started/4_benchmarks.html#inverse-kinematics
I also noticed that you changed the example to only run the for loop once. You need to warmup the ik solver on the gpu by running a few times (3) before timing. We hence use a for loop of 10.
I also noticed that you changed the example to only run the for loop once. You need to warmup the ik solver on the gpu by running a few times (3) before timing. We hence use a for loop of 10.
Hi Balakumar, the time consumption did change after adding warm up.
As we implement our own robotic arm example, what else should we pay attention to? For example, warm up the process as you just mentioned
By the way, if we want to share our robot as an example with you, what do we need to provide?
You need to provide the urdf, meshes and the robot configuration file.
If it’s not a bug, please use discussions: https://github.com/NVlabs/curobo/discussions
Please provide the below information in addition to your issue:
Hi @balakumar-s , I just execute the example. But the time cost for franka IK example is strange and does not math the time mentioned in the paper.
Result
Code