Update README.md
Browse files
README.md
CHANGED
|
@@ -37,7 +37,8 @@ High-fidelity FP8 quantization for diffusion models (Z Image). HSWQ uses **sensi
|
|
| 37 |
|
| 38 |
| Feature | V1: Standard Compatible | V2: High Performance Scaled |
|
| 39 |
| :--- | :--- | :--- |
|
| 40 |
-
| **Compatibility** | Full (100%), any FP8 loader |
|
|
|
|
| 41 |
| **File format** | Standard FP8 (`torch.float8_e4m3fn`) | Extended FP8 (weights + `.scale` metadata) |
|
| 42 |
| **Image quality (SSIM)** | ~0.89 (theoretical limit) | ~Unable to measure at this time |
|
| 43 |
| **Mechanism** | Optimal clipping (smart clipping) | Full-range scaling (dynamic scaling) |
|
|
|
|
| 37 |
|
| 38 |
| Feature | V1: Standard Compatible | V2: High Performance Scaled |
|
| 39 |
| :--- | :--- | :--- |
|
| 40 |
+
| **Compatibility** | Full (100%), any FP8 loader | The scaled model does not perform well in the current ComfyUI.
|
| 41 |
+
|
|
| 42 |
| **File format** | Standard FP8 (`torch.float8_e4m3fn`) | Extended FP8 (weights + `.scale` metadata) |
|
| 43 |
| **Image quality (SSIM)** | ~0.89 (theoretical limit) | ~Unable to measure at this time |
|
| 44 |
| **Mechanism** | Optimal clipping (smart clipping) | Full-range scaling (dynamic scaling) |
|