ybgdgh / co-navgpt Goto Github PK

Hi, thank you for your work, I really appreciate it.

Could you provide me the command for downloading the dataset as mentioned in "Setting up datasets"?

or can you provide some detail instructions for constructing them?

Thank you for sharing the codes.

Can you provide the docker image file that you used?

When I try to install scikit-fmm==2019.1.30 using 'pip', I encounter an issue. I suspect this may be due to the version of numpy. The numpy version in my environment is 1.26.3, which was automatically installed when I installed torchv1.12.1. Can you please provide the recommended version of numpy?"

the error is :

After run code: python -m habitat_sim.utils.datasets_download --username xxx --password xxx --uids hm3d_minival_v0.2
I got: Data download failed, no datasource named hm3d_minival_v0.2

After removing "_v0.2", it works. But the dataset structure is totally different from yours
data/scene_datasets
data/versioned_data

Therefore, i dont know what to do next. Could you give me some good advice, please?

Have you ever used this in practice? Or is it feasible in practice?

Your work is very awesome work and I'm curious is it a solution for the 2023 Challenge?

Hi! Thank you for your work!

I encountered problems when i am doing python setup.py install for habitat-sim after git checkout to tags/challenge-2022. The error log:

Install the project...
[ 24%] Built target corrade
[ 24%] Built target MagnumSceneGraph
[ 24%] Built target MagnumShaderTools
-- Install configuration: "RelWithDebInfo"
-- Installing: /usr/local/lib/libz.so.1.2.11
CMake Error at cmake_install.cmake:65 (file):
file INSTALL cannot copy file
"/home/xxxx/habitat-sim/build/deps/openexr/zlib-build/libz.so.1.2.11"
to "/usr/local/lib/libz.so.1.2.11": Permission denied.

Could you give some suggestions on that?
Thank you!

How do I generate object_norm_inv_perplexity.npy?

Hi, could you please explain how you use mapping constants in the constants.py? I noticed the code use coco dataset for prediction, but it also use mp_categories_mapping array to remapping the segmentation result in line 605(llm_agents.py). I am currently run the code in my own environment for different object categories, it will be help a lot if you could explain this.

          you can refer this [repo](https://github.com/MIT-SPARK/llm_scene_understanding/blob/main/data_generator.py) to generate this file based on your dataset by yourself. Hope this can help you

Originally posted by @ybgdgh in #10 (comment)

After run code: python -m habitat_sim.utils.datasets_download --username xxx --password xxx --uids hm3d_minival_v0.2
I got: Data download failed, no datasource named hm3d_minival_v0.2

After removing "_v0.2", it works. But the dataset structure is totally different from yours
data/scene_datasets
data/versioned_data

Can this download file from the latest version repository in your reply download the dataset structure that is exactly the same as yours? I don't know how to use this download file.

I want to obtain the dataset structure that is exactly the same as in Co-NavGPT. Could you please give me some suggestions? Thank you very much!

Have you ever meet this error?when I use the command "python setup.py install --headless

"

Hi, thank you for the work provided!

I was currently working on a multi-robot search task, and your implementation for the multi-robot setting helps a lot. I noticed that in update_metric function of DistanceToGoal(Measure) of nav.py in multi-robot-setting/habitat folder, the self._metric = min(distance_to_target) is within the loop of for i in range(self._sim.habitat_config.NUM_AGENTS):. I think it should be outside, same level to the for loop.

In addition, i am curious about why the ''SPL" metric did not include 2 agents' information, as the reset_metric in SPL class, self._previous_position = self._sim.get_agent_state().position will defaultly get the agent0 information only.

Best regards

I am interested in following your excellent work. However, I have encountered some issues.

The results of the Co-Nav are as follows, which are lower than the results reported in the paper (success: 0.661):

When I run the code, I encountered a bug in the function parse_answer. I observed that the output of ChatGPT contains an extra newline character. So I add the code: if 'robot' in line. Does it affect the results?

Have you ever meer this error?When I use the command "python setup.py install --headless

"