AkashicPulse EQVAE is a model based on AkashicPulse series, but fine-tuned to utilize Anzhc's EQVAE as its VAE. (In theory,) This improve the model generation color and accuracy.
Recommended settings [UPDATED FOR V1.0.x]:
Same like v0.2, the model is in between needing RescaleCFG and not needing it. Expect to turn on and off this a lot.
The usage of artist tags IS HIGHLY RECOMMENDED. Put it at the FRONT of your prompt (and use BREAK if your generation platform support it).
Steps: 32-50
CFG: 7-10 with RescaleCFG on or off, or 4.5-6 if you don't wanna touch RescaleCFG
RescaleCFG: 0.7 OR OFF, inspect generated image quality if something is off.
Sampler: Euler a (or try AkashicSolver available on Adept Sampler with AkashicAOS).
Noise Scale: 0.9, using 1.0 will cause the model to over-generate noise.
Recommended negative prompt:
worst quality, low quality, bad anatomy, bad hands, text, signature, watermark, username, artist name, [things you want to remove]WARNING: Using excessive negative prompt might resulting the model to lose its quality.
Model Specs:
v1.0.2
Based on: AkashicPulse EQVAE v1.0.1-Base
Dataset size: ~21k
Epoch posted: Epoch 10 out of 10
EQVAE used: Anzhc's EQVAE B7
v1.0-v1.0.1
Based on: AkashicPulse EQVAE v0.2
Dataset size: ~4.6k
Epoch posted: Epoch 15 out of 15
EQVAE used: Anzhc's EQVAE B7
v0.2 [CHECK ABOUT THIS VERSION FOR DETAILS]
Based on: AkashicPulse EQVAE v0.1
Dataset size: ~9.4k
Epoch posted: Epoch 7 out of 10.
EQVAE used: Anzhc's EQVAE B7
v0.1
Based on: AkashicPulse 4.1
Dataset size: 10.6k, repeated twice.
Epoch posted: Epoch 9 out of 10.
EQVAE used: Anzhc's EQVAE B7
Description
Long story short, using the correct epoch and newer knowledge transfer methods. Imagine v1.0.2 but better, and no early access.
FAQ
Comments (4)
error: TypeError: linear(): argument 'weight' (position 2) must be Tensor, not NoneType
Some reports on my DM saying that there's a problem with v1.0.3 on Comfy. Tbh I didn't check Comfy and was checking the checkpoint with SD WebUI Forge Neo. EDIT: I didn't find the same issue. Check comment for details.
I didn't find the same issue on my comfy installation. https://imgur.com/a/t8HgPV5
If you happen to have issues, please make sure your comfy installation is both using the latest version AND the latest dependencies. If it still happening, please post your workflow and I'll try to help you as much as I could.
Here is some information that might be helpful: in the latest version of ComfyUI, running v1.0.3 triggers a RuntimeError: mat1 and mat2 shapes cannot be multiplied (77x768 and 3072x768) during the CLIP stage. I managed to get it working and achieving excellent results by patching the CLIP from the previous version into 1.0.3. I’ve also cross-referenced the SHA256 hashes to ensure the downloaded model isn't corrupted.