Hi,
Congrats for the good work! However, I found the released code miss an important component in the paper.
The paper proposes to generate staged cost functions using VLM and Hamster. However, I found that this code base generates cost functions using privileged information return from the simulator (see this line where keypoints are determined by the target object pose.
I'm wondering if this is the exact code base for conducting the experiments shown in the paper.
Hi,
Congrats for the good work! However, I found the released code miss an important component in the paper.
The paper proposes to generate staged cost functions using VLM and Hamster. However, I found that this code base generates cost functions using privileged information return from the simulator (see this line where keypoints are determined by the target object pose.
I'm wondering if this is the exact code base for conducting the experiments shown in the paper.