The pre-training model

[sunkaifan]

Can you provide the pre-training model of spanning graph prior and layout decoder? Thank you very much！

[arjuntheprogrammer]

I have created my own checkpoints by training them on Nvidia L4 GPU.

For 2000 epochs (living room) and:

• for layout decoder it took 2 days
• for scene graph it took 4 days

Let me know what would be the best way to include the checkpoints with this repo for everyone to use.

[arjuntheprogrammer]

Hi @sunkaifan - I have created a PR on Hugging face and uploaded the checkpoints for SceneGraph and LayoutDecoder for bedroom.

Please check: https://huggingface.co/datasets/chenguolin/InstructScene_dataset/tree/cb526c11234b8e6ad76a49f9754e236a51b28c24

https://huggingface.co/datasets/chenguolin/InstructScene_dataset/tree/f658bd29303172cf84ac6fa8b625e83ded56ec3a

[sunkaifan]

Thank you, very good job! Will the pre-training model of living room and dining room be open source in the future?

[arjuntheprogrammer]

Thank you, very good job! Will the pre-training model of living room and dining room be open source in the future?

If I get time, I will upload it for the living and dining room as well.

[sunkaifan]

Thanks for your reply!

[chenguolin]

Hi @arjuntheprogrammer, thank you very much for sharing the pretrained parameters! :)

I've updated README to inform that the pretrained models are uploaded to HuggingFace.