Spaces:
Runtime error
Car model
Hi - how was the car model trained - ESD-x or ESD-u ?
Hi, its esd-u
could you provide parameters used for finetuning (if they are different for default e.g. 1000 iterations)?
The models are created with the current code (developmental) using diffusers training. Here we train them for 150 epochs and with default parameters. You can clone this current Huggingface repo and use ESD-u with 150 epochs.
i was thinking that one can use https://github.com/rohitgandikota/erasing/blob/main/train-scripts/train-esd.py for default 1000 iterations, this was not trained this way?
For sure! I would suggest default iterations of 1000, but depending on your preference and use case, you can play around with the iterations 500 and negative guidance of 1 too.
The current model that was provided in the gradio demo is from a different codebase (developmental phase) that is provided here https://huggingface.co/spaces/baulab/Erasing-Concepts-In-Diffusion/tree/main
Ah I see, thank you will test it then :)