Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

An error raised by tyro #1

Open
DJNing opened this issue Jul 11, 2024 · 5 comments
Open

An error raised by tyro #1

DJNing opened this issue Jul 11, 2024 · 5 comments

Comments

@DJNing
Copy link

DJNing commented Jul 11, 2024

Thanks a lot for the great work! Yet I encourtered some problems while runing the demo

Need help for the error for the configuration loading

I am trying to run the test based on the code. Yet, I found one error in:

AllConfigs = tyro.extras.subcommand_type_from_defaults(config_defaults, config_doc)

It says the input arg for defaults (config_defaults) should have len(defaults) > 2.

While I am not familiar to tyro, could you guys tell a bit on how to solve this?

Extra dependency for the repo:

pip install timm jaxtyping omegaconf typeguard colorama

Minor change for the mamba_ssm:

For the latest version of mamba_ssm (2.2.2) the following line:

from mamba_ssm.ops.triton.layernorm import RMSNorm, layer_norm_fn, rms_norm_fn

should be changed to from mamba_ssm.ops.triton.layer_norm import RMSNorm, layer_norm_fn, rms_norm_fn

@DJNing
Copy link
Author

DJNing commented Jul 11, 2024

A quick fix for the code with latest tyro.

Change the following line:

AllConfigs = tyro.extras.subcommand_type_from_defaults(config_defaults, config_doc)

to AllConfigs = Options

@florinshen
Copy link
Collaborator

thanks for your detailed solution, we will update it into our env requirements

@hgdzhx
Copy link

hgdzhx commented Sep 5, 2024

After the correction in DJNing, some bugs were fixed, but later some CUDA functions were found to be unusable (such as xFormers). After installing xFormers with pip, the system uninstalled dcuda11.8+torch 2.1.0 and updated it to nvidia-nccl-cu12+torch 2.4.0. There are quite a few questions and I don't know how to operate it. Could you please give me some guidance

@florinshen
Copy link
Collaborator

sure, hgdzhx. xformers actually is not neccessary during inference in current code, so you can ignore the installation of xformers.

If you want to install xformer for speedup, a compromise is to build the xformers from source.

@hgdzhx
Copy link

hgdzhx commented Sep 13, 2024

I sincerely appreciate your reply. In the follow-up repairing and reviewing of the code, I found that the open source code can only support fixed camera parameters of the ''S3'' format image input. So can you open the complete code? I find your work very innovative and will keep an eye on you. Hope you have a nice day.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants