Files changed (1) hide show
  1. README.md +4 -3
README.md CHANGED
@@ -5,6 +5,9 @@ tags:
5
  - SVDQaunt
6
  - SVDQ
7
  - Quantization
 
 
 
8
  language:
9
  - en
10
  base_model_relation: quantized
@@ -19,8 +22,6 @@ Model Files
19
  - [svdq-int4_r32-Jib-Mix-Flux-V12.safetensors](#): SVDQuant INT4 (rank 32) Jib Mix Flux V12 model. For users with non-Blackwell GPUs (pre-50-series).
20
  - [svdq-fp4_r32-Jib-Mix-Flux-V12.safetensors](#): SVDQuant NVFP4 (rank 32) Jib Mix Flux V12 model. For users with Blackwell GPUs (50-series).
21
 
22
-
23
-
24
  ### Quality Evaluation
25
 
26
  Below is the quality and similarity evaluated with 256 samples from MJHQ-30K dataset. (BF16 is the unqauntized model. INT W4A4 is INT4 and NVFP4 is FP4)
@@ -28,7 +29,7 @@ Below is the quality and similarity evaluated with 256 samples from MJHQ-30K dat
28
  | Model | Precision | Method | FID | IR | LPIPS | PSNR |
29
  |----------------------------|-----------|-----------|--------------------|-----------------|----------------------|-------------------|
30
  | Jib Flux Mix V12 (25 steps)| BF16 | -- | 170.37 | 0.946 | -- | -- |
31
- | | INT W4A4 | SVDQ | 135.50 | 0.795 | 0.255 | 20.64 |
32
  | | NVFP4 | SVDQ | 170.99 | 0.955 | 0.220 | 20.71 |
33
 
34
  If you find these useful please consider: <a href='https://ko-fi.com/B0B21MPRDT' target='_blank'><img height='36' style='border:0px;height:36px;' src='https://storage.ko-fi.com/cdn/kofi6.png?v=6' border='0' alt='Buy Me a Coffee at ko-fi.com' /></a>
 
5
  - SVDQaunt
6
  - SVDQ
7
  - Quantization
8
+ - jib
9
+ - j1b
10
+ - Jib Mix Flux
11
  language:
12
  - en
13
  base_model_relation: quantized
 
22
  - [svdq-int4_r32-Jib-Mix-Flux-V12.safetensors](#): SVDQuant INT4 (rank 32) Jib Mix Flux V12 model. For users with non-Blackwell GPUs (pre-50-series).
23
  - [svdq-fp4_r32-Jib-Mix-Flux-V12.safetensors](#): SVDQuant NVFP4 (rank 32) Jib Mix Flux V12 model. For users with Blackwell GPUs (50-series).
24
 
 
 
25
  ### Quality Evaluation
26
 
27
  Below is the quality and similarity evaluated with 256 samples from MJHQ-30K dataset. (BF16 is the unqauntized model. INT W4A4 is INT4 and NVFP4 is FP4)
 
29
  | Model | Precision | Method | FID | IR | LPIPS | PSNR |
30
  |----------------------------|-----------|-----------|--------------------|-----------------|----------------------|-------------------|
31
  | Jib Flux Mix V12 (25 steps)| BF16 | -- | 170.37 | 0.946 | -- | -- |
32
+ | | INT W4A4 | SVDQ | 171.64 | 0.907 | 0.253 | 20.31 |
33
  | | NVFP4 | SVDQ | 170.99 | 0.955 | 0.220 | 20.71 |
34
 
35
  If you find these useful please consider: <a href='https://ko-fi.com/B0B21MPRDT' target='_blank'><img height='36' style='border:0px;height:36px;' src='https://storage.ko-fi.com/cdn/kofi6.png?v=6' border='0' alt='Buy Me a Coffee at ko-fi.com' /></a>