ram81 / goat-bench Goto Github PK

Hi!
Thanks for your fantastic work.
I just want to run the evaluation part to see how the benchmark work for GOAT task. Is it possible to offer the checkpoint, and the caches which have been mentioned at the config documents (object_cache, image_cache, language_cache)?

Hi, thanks for the wonderful work!

I have a question about the Performance of individual skills used for RL Skill Chain baseline which is shown in Table5 in the Appendix.
Is VAL UNSEEN EASY referring to Val Seen Synonyms Split or a subset of the Val Unseen Split?

Looking forward to your reply!

Hi,

Thank you for your excellent work. Could you please tell me if the paper is available to the public?

Thanks for the great work!!

In the script of "1-languagenav-eval.sh" I saw a path of "DATA_PATH="data/datasets/languagenav/hm3d/v5_final/". Seems like there is pure language nav dataset that only contains the language dataset. I wonder will this dataset be released.

Or maybe put it in this ways, is there other way to let the agent be only trained/evaluated with language description as the goal, rather than with different types in sequence.

When I run ,SenseAct-NN Monolithic Baseline'Evaluation ,I get the following error.I have configured the environment according to the instructions and executed 'pip install -e .'

and the link contains several datasets.
Is the largest dataset the "full dataset" mentioned?

Hi, thanks for the great work. Could you please provide the pretrained policy weights?

Thank you for your work!

I tried to download with the following command from here, but it did not work.

python -m habitat_sim.utils.datasets_download --username <api-token-id> --password <api-token-secret> --uids hm3d_minival_v0.2

I believe the repo of the dataset should be here, and I have downloaded hm3d-minival-habitat-v0.2.tar, hm3d-minival-semantic-configs-v0.2.tar, hm3d-train-habitat-v0.2.tar, hm3d-val-habitat-v0.2.tar. Neither of these has the structure like this:

├── goat-bench/
│  ├── data
│  │  ├── scene_datasets/
│  │  │  ├── hm3d/
│  │  │  │  ├── JeFG25nYj2p.glb
│  │  │  │  └── JeFG25nYj2p.navmesh

Thanks for your appealing work on the multi-modal lifelong navigation project. I encountered an issue when trying to run the provided scripts.

The error I'm seeing is:

AttributeError: 'BaseEpisode' object has no attribute 'tasks

This error occurs in the following code:

I found that the current_episodes_info variable is a list of BaseEpisode objects, whose source code is here:
https://github.com/facebookresearch/habitat-lab/blob/2c7519b1baead8d5bc9557dd53d2a01e4aa9c6e5/habitat-lab/habitat/core/dataset.py#L43-L53

It seems the BaseEpisode class does not have an attribute named tasks, which is causing the error.

I was able to work around this by commenting out the lines of code that were trying to access the tasks attribute. However, this is not an ideal solution, as it may indicate a deeper issue with the code.

Could you please take a look and provide some guidance on how to resolve this error? I'm happy to provide any additional information that might be helpful.

Thank you for your time and great work on this project!

I installed the packages according to the README, but encountered some issues.

I needed to additionally install the lavis and vc_models packages. Also, even after installing the above packages, I still receive the following error when I try to run the train script.

MissingConfigException: In 'ver_goat.yaml': Could not load 'habitat_baselines/rl/policy/obs_transforms/resize'.

Is v0.2.3 the correct version of habitat-lab?

If you have the correct installation procedure with modifications, could you let me know?
I look forward to your response.