Dockerfile

[detrin]

Hello, would you consider adding docker file with instructions to the model? That would make it much easier for people to use this project locally.

[ftaghiyev]

Hi @detrin ,

Thanks for raising this! I’m currently working on a Dockerfile and short usage instructions (assuming the authors are okay with it). I’ll update this issue once I have something ready to test.

[yt-koike]

@detrin @ftaghiyev Hello there!

I've opened a PR introducing a Dockerfile for ml-sharp, along with a Gradio-based web UI to enhance usability.
I would appreciate any comments or feedback.

Thank you.

[frostworx]

thank you very much @yt-koike!

I just gave it a quick try and it worked fine.

Sharp command took 299.684 seconds

with a GeForce RTX 5060 Ti, so I guess it might have used the CPU.

this might be a problem on my end though and needs further testing (restructuring that system currently)

edit:
tried the command from within the docker using the teaser.jpg and it required just a few seconds.

edit2:

just gave it another try with your ui and it worked excellent within a few seconds!
Thank you very much again! :)

[yt-koike]

@frostworx Well spotted! I think it took about five minutes at the first inference because of compiling. When I was testing the docker image in my environment, I've faced the same problem: the slow start problem. I checked the process and then noticed cicc and ptxas were working. Thus I think that the compile process takes a bit of time at the first start as indicated in README.md; The gsplat renderer takes a while to initialize at the first launch. In my environment, it took about 15 minutes, but after that Sharp worked quite fast.

I'm glad to hear that it worked on your environment too! Enjoy :)