Hi @ajwagen, thanks for your great work! I am trying to reproduce the experiments in the paper.
I am currently trying to reproduce the transport experiment with python train_dsrl.py --config-path=cfg/robomimic --config-name=dsrl_transport.yaml and this is the result.
The experiment has been running for four days on a RTX 5090 GPU.
Also, I was curious to check the performance of the NA version of the algorithm, just by setting dsrl_na in the previous config file. And this is the result.
Both plots are quite different from the result presented in Fig.4 of the paper. Am I missing anything or doing something wrong?
Thanks in advance!
Hi @ajwagen, thanks for your great work! I am trying to reproduce the experiments in the paper.
I am currently trying to reproduce the transport experiment with
python train_dsrl.py --config-path=cfg/robomimic --config-name=dsrl_transport.yamland this is the result.The experiment has been running for four days on a
RTX 5090GPU.Also, I was curious to check the performance of the NA version of the algorithm, just by setting
dsrl_nain the previous config file. And this is the result.Both plots are quite different from the result presented in Fig.4 of the paper. Am I missing anything or doing something wrong?
Thanks in advance!