Delete fp8_scaled_kj
Browse files
fp8_scaled_kj/README.md
DELETED
|
@@ -1,11 +0,0 @@
|
|
| 1 |
-
Better fp8 scaled models based on quantization code from https://github.com/Tencent-Hunyuan/HunyuanVideo/blob/main/hyvideo/modules/fp8_optimization.py
|
| 2 |
-
|
| 3 |
-
Works in both wrapper (latest version) and native Wan video workflows.
|
| 4 |
-
|
| 5 |
-
14B-T2V comparison test without LoRAs, 25 steps, 832x480x81
|
| 6 |
-
---
|
| 7 |
-
|
| 8 |
-
<video controls autoplay width=50% src=https://cdn-uploads.huggingface.co/production/uploads/63297908f0b2fc94904a65b8/o6sd_sDb47pnBPxshyypA.mp4></video>
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
fp8_scaled_kj/Wan2_1-T2V-14B_fp8_e4m3fn_scaled_KJ.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:5519e566e620037b1adb399886143991036d27d44455f41190410967a2fc130d
|
| 3 |
-
size 14526876890
|
|
|
|
|
|
|
|
|
|
|
|