Update README.md
Browse files
README.md
CHANGED
|
@@ -15,6 +15,11 @@ Separated LTX2.3 checkpoint for alternative way to load the models in Comfy
|
|
| 15 |
|
| 16 |
The fp8 quantizations were done with the basic static weight scales and are set to not run with fp8 matmuls, the models marked `input_scaled` additionally have activation scaling, and are set to run with fp8 matmuls on supported hardware.
|
| 17 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
Tiny VAE by [madebyollin](https://github.com/madebyollin/taehv/)
|
| 19 |
|
| 20 |
Can be used like this currently:
|
|
|
|
| 15 |
|
| 16 |
The fp8 quantizations were done with the basic static weight scales and are set to not run with fp8 matmuls, the models marked `input_scaled` additionally have activation scaling, and are set to run with fp8 matmuls on supported hardware.
|
| 17 |
|
| 18 |
+
As this is first time I'm attempting to do calibrate input scales, these are pretty experimental, but result wise seems to work, this is a test on a 4090, 8 steps with distill:
|
| 19 |
+
|
| 20 |
+
<video controls autoplay width=50% src=https://cdn-uploads.huggingface.co/production/uploads/63297908f0b2fc94904a65b8/ALNr3_j0klp29fHkI3pyt.mp4></video>
|
| 21 |
+
|
| 22 |
+
|
| 23 |
Tiny VAE by [madebyollin](https://github.com/madebyollin/taehv/)
|
| 24 |
|
| 25 |
Can be used like this currently:
|