Versions
int8: recommended. Fast, accurate, compatible with almost any GPU.
mxfp8: added for comparison. In theory (and according to nVidia PR) should be more accurate than int8, but in practice I was not able to spot any definitive advantages. A bit slower than int8, but still faster than original bf16. Compatible only with RTX 50xx series (Blackwell).
Performance on my setup
original bf16 (baseline): 2.20 it/s +0%
int8: 3.23 it/s +46%
int8 + torch compile (comfy core): 3.59 it/s +63%
int8 + turbo lora, cfg=1: 6.50 it/s +295%
int8 + turbo lora, cfg=1 + torch compile (comfy core): 7.55 it/s +343%
mxfp8: 2.58 it/s +17%
This is high quality int8 quantized version of base Anima v1.0 model. It retains ~90% of original model quality, but uses about 50% less VRAM and also runs faster on almost any nVidia GPU (AMD not tested). Nice trade-off, especially for low-end GPUs.
Can be used as a drop-in replacement for original Anima model in latest ComfyUI, no custom nodes required. If you have troubles running the model make sure that you updated both ComfyUI itself and its dependencies (e.g.pip install -U -r requirements.txt on manual linux install).
Converted to int8 / mxfp8 using convert_to_quant script.
Description
Anima-Base v1.0.
int8, ConvRot group size 256, rowwise, learned rounding SVD
Proper, fast version.
FAQ
Comments (15)
Hi I ran into this error:
Server error: 'int8_tensorwise'
I'm using Krita AI plugin though, you you happen to know a fix for this?
You have to update both ComfyUI itself and its dependencies (pip install -U -r requirements.txt on manual linux install). Have you done this?
Comfy fails to load loras with it
Loras are working on my setup. Have you updated Comfy and libs to latest version (see model description)?
The INT8 saves about 10-15 seconds on my setup I didn’t notice a significant loss of quality, everything, as always, depends on the prompt, sampler/scheduler and parameters - but overall the model feels lighter and faster. Thanks for sharing.
hi would you mind to make wai-anima v1 too?
Please post some comparison images showing the quality and the generation time.
I posted a couple of comparison images.
int8 generation is ~45% faster on my setup, but effect varies a lot on different GPUs.
I've already run multiple rounds of testing. The image quality is on par with FP8, with almost no visible difference—possibly no difference at all. But it's more than 10~30% faster.
@somedoby Thanks
@BIG_A Thanks for the info.
@ikiru99percent BTW，Please update your CUDA version to 13.0. Otherwise, it may still be slow.
@BIG_A My CUDA version is 13.2 [rtx 5060 ti]
@ikiru99percent that s ok