AmoPulse is a LoKr full-matrix fine-tune based on AkashicPulse.
The backstory for this model is simple: full fine-tuning is expensive, and I am no longer making a profit on AI. This is the most cost-effective way to perform a semi-full fine-tune without significantly compromising model quality.
Changelog:
v1: Trained a LoKr on top of AkashicPulse v4.1 using a 4.6k Hololive Indonesia dataset. This improves AkashicPulse's ability to generate Hololive Indonesia characters.
The model falls under Fair AI Public License 1.0-SD with no additional terms.
Description
Initial release
FAQ
Comments (1)
ASk, this model is nerf version of AkashicPulse or upgrade with additional training on Hololive Indonesia dataset?