add web demo/model to Huggingface about cvpr2022-dagan HOT 18 CLOSED

The demo is currently running demo_dagan script every inference call, which means the model is reloaded every time and this is what might be making things slow.

from cvpr2022-dagan.

That's awesome, thanks, I will Integrate DaGAN into Huggingface Spaces in near future.

from cvpr2022-dagan.

Hi @AK391 ,

The Hugging Face is a great platform for us to present our project. However, I met a problem when deploying my code on it, It seems quite simple for you experienced guys:

Could you help me with this problem?

from cvpr2022-dagan.

@harlanhong Hi, it looks like the requirements.txt file is missing so similar to github you can add a requirements.txt file to spaces https://github.com/harlanhong/CVPR2022-DaGAN/blob/master/requirements.txt

from cvpr2022-dagan.

@harlanhong Hi, it looks like the requirements.txt file is missing so similar to github you can add a requirements.txt file to spaces https://github.com/harlanhong/CVPR2022-DaGAN/blob/master/requirements.txt

Thanks, it works. I will share the space link with you soon.

from cvpr2022-dagan.

How to use GPU in spaces. I can now run my code on it but it cost too much time when using the CPU.
When I load my model on GPU, an error will occur:

from cvpr2022-dagan.

@harlanhong Hi, from the spaces docs: https://huggingface.co/docs/hub/spaces

Each Spaces environment is limited to 16GB RAM and 8 CPU cores. Organization subscribers (Lab, Startup, and Enterprise) can access Spaces with one T4 GPU on a case-by-case basis. Please email us at website at huggingface.co or let us know on Twitter if you need one.

I can also send a request to assign a gpu for this space

from cvpr2022-dagan.

@harlanhong Hi, from the spaces docs: https://huggingface.co/docs/hub/spaces

Each Spaces environment is limited to 16GB RAM and 8 CPU cores. Organization subscribers (Lab, Startup, and Enterprise) can access Spaces with one T4 GPU on a case-by-case basis. Please email us at website at huggingface.co or let us know on Twitter if you need one.

I can also send a request to assign a gpu for this space

Many thanks.

from cvpr2022-dagan.

Hey there! 👋 We opened an internal PR to provide a temporal GPU, we'll let you know once it's deployed.

from cvpr2022-dagan.

thanks @osanseviero also @harlanhong huggingface can also host models for free, so the pretrained checkpoints in one drive for example can be added to the huggingface model hub under your username or a organization name, more details here: https://huggingface.co/docs/hub/adding-a-model

from cvpr2022-dagan.

This Space has GPU now! 🚀 🚀

from cvpr2022-dagan.

thanks @osanseviero, was testing out the space, still seem to be taking some time, 200 seconds so far with one of the examples and the output is not showing @harlanhong

from cvpr2022-dagan.

The demo is currently running demo_dagan script every inference call, which means the model is reloaded every time and this is what might be making things slow.

Oh, I see. Thanks for your reminder.

from cvpr2022-dagan.

thanks @osanseviero, was testing out the space, still seem to be taking some time, 200 seconds so far with one of the examples and the output is not showing @harlanhong

Hi @AK391, I have modified my code to load everything before launching, but when I test it I get an error, however, the log file doesn't report anything about that error.

from cvpr2022-dagan.

@harlanhong sorry for the late response, are you still getting this error and how long is inference taking on the gpu space?

from cvpr2022-dagan.

@harlanhong currently its running over 600 seconds

from cvpr2022-dagan.

@harlanhong any update on this?

from cvpr2022-dagan.

Hi @harlanhong! Would you be interested in making the checkpoint available on the Hub as a model repo? It would make it easy for our community to find your model and you could attach some tags and documentation to it, like this model: https://huggingface.co/bert-base-uncased

There's documentation on how to do this, but I'm also happy to take it on or help out!

from cvpr2022-dagan.