mees / hulc2 Goto Github PK

Thank you very much for sharing your great work, but I found some errors when running the evaluate_policy.py on your pretrained checkpoints as indicated in the README.md, and I'm not sure whether it's a bug or some extra configureations should be done before running the code:

1. The PlayDataModule can't load the taco-robot downloaded on kaggle because of the different path structure, nor could the .hydra folder and embeddings.npy be found in the dataset
2. The file "config/hulc2/rollout/aff_hulc2.yaml" could not be found while it's required in #hulc2/evaluation/manager_aff_lmp.py, has it been renamed to another file?

Great work! It's very impressive, but I am wondering if there is any possibility that you could provide pre-trained models specifically trained on the CALVIN dataset. On CALVIN website, HULC++ is current SOTA, could you please provide that version of model?

Thanks so much!

Thank you for your excellent work!
I would like to kindly ask if you have the checkpoint of HULC++ in ABC->D settings. Alternatively, do you have any results from HULC++ in the ABC->D setting?

Hi, thanks for sharing your codebase for great work!

I'm trying to reproduce HULC++'s evaluation results in CALVIN.
I found that the validation loss of depth prediction increases very highly compared to the training's one in affordance model training, as shown in the figure below.
It'd be very helpful if you can confirm whether this is an expected behavior, or if I'm missing something while following your instructions.

I actually encountered very suspicious error during the final stage of processing validation data.
The error occurred in this line as add_norm_values tries to indexing data['training'] which appears to be reasonably missing while processing validation data.
Would this error have affected data sanity?

Thanks!

python hulc2/evaluation/evaluate_policy.py --dataset_path dataset/task_D_D --train_folder real_world_checkpoints/aff_model_single/checkpoints
Global seed set to 0
Traceback (most recent call last):
File "/home/systemtec/hulc2/hulc2/evaluation/evaluate_policy.py", line 94, in
main()
File "/home/systemtec/hulc2/hulc2/evaluation/evaluate_policy.py", line 81, in main
checkpoints = [Path("epoch=%s.ckpt" % int(chk)) for chk in args.checkpoints.split(",")]
AttributeError: 'NoneType' object has no attribute 'split'

Could you provide a Docker image of the engineering environment? We have read HULC2's paper and are very interested in your proposed methods and ideas. However, some problems were encountered during the reproduction process. Therefore, may I ask if it is possible to provide a Docker code environment? Thank you very much!

Could you provide a Docker image of the engineering environment? We have read HULC2's paper and are very interested in your proposed methods and ideas. However, some problems were encountered during the reproduction process. Therefore, may I ask if it is possible to provide a Docker code environment? Thank you very much!