Questions about training configs

by ssonpull519 - opened Nov 13, 2023

Nov 13, 2023

Thank you for sharing your interesting work with the community.

Can we get a training details and configs for LCM-LoRA models, such as training dataset / LoRA initialization / learning rate / optimizer / etc.?

Thanks!

sayakpaul

Nov 13, 2023

Please refer to the paper for that:
https://arxiv.org/abs/2311.05556

ssonpull519

Nov 13, 2023

@sayakpaul I know there is a report but is training config for LCM-LoRA the same as the one used in LCM?
Also I recently noticed that diffusers have implemented LCM including training script here, can I refer to this for reproducibility of your report?

sayakpaul

Nov 13, 2023

The report I linked to is for LCM-LoRA which is what you asked for if I am not mistaken?

Training config for LCM-LoRA is not the same one as LCM SDXL.

Yes, you can refer to the scripts for reproducibility. For more details, I defer to @valhalla for detailing the model cards more.

ssonpull519

Nov 13, 2023

•

edited Nov 13, 2023

@sayakpaul Yeah it does not provide details of training for LCM-LoRA but sufficient for me based on your response anyways now. Thank you for your clarification!

ssonpull519 changed discussion status to closed Nov 14, 2023

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment