search

dyson-ai / hdp

[CVPR 2024] Hierarchical Diffusion Policy for Multi-Task Robotic Manipulation
MIT License
117 stars 8 forks source link

Can not run the evaluation code due to missing of PerAct checkpoint #6

Closed andvg3 closed 3 weeks ago

andvg3 commented 3 weeks ago

Hi authors,

I am trying to run the evaluation code using your instruction:

python eval.py rlbench.tasks="[<your task>]" rlbench.headless=False method.model_path=<path to rk-diffuser ckpt> framework.logdir=<path to peract ckpt dir>

As my guessing, <path to rk-diffuser ckpt> refers to the checkpoint of the RK-Diffuser model, which locates in the snapshots folder and <path to peract ckpt dir> refers to the checkpoint in the PerAct. Nevertheless, it seemed not working and I encoutered the following error:

Error executing job with overrides: ['rlbench.tasks=[reach_target]', 'rlbench.headless=False', 'method.model_path=snapshots/low_level/reach_target/2024-08-28_17:36:56/model_10000.pt', 'framework.logdir=saved_models/ckpts/multi/PERACT_BC/seed5'] Traceback (most recent call last): File "/home/XXX/hdp/eval.py", line 253, in main raise Exception("Missing seed%d/config.yaml" % start_seed) Exception: Missing seed0/config.yaml Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.

Of course, I have double checked to ensure the train_config_path exists as a relative path. I also tried to change to absolute path but still, it did not work.

I hope you can help me with this issue. Thank you very much in advance!

Best regards, An Vuong

andvg3 commented 3 weeks ago

Oh, never mind, just change the relative path to the absolute path works for me.