Clarify the GDN/KDA operator and target GPU in the title
Browse files
README.md
CHANGED
|
@@ -11,7 +11,7 @@ tags:
|
|
| 11 |
- linear-attention
|
| 12 |
---
|
| 13 |
|
| 14 |
-
# LeapQuant
|
| 15 |
|
| 16 |
One repository containing the Gated DeltaNet (GDN) and Kimi Delta Attention
|
| 17 |
(KDA) window-boundary state re-quantization kernels. Both use CuTe DSL and
|
|
|
|
| 11 |
- linear-attention
|
| 12 |
---
|
| 13 |
|
| 14 |
+
# LeapQuant GDN/KDA State Requantization — CuTe DSL on NVIDIA B300
|
| 15 |
|
| 16 |
One repository containing the Gated DeltaNet (GDN) and Kimi Delta Attention
|
| 17 |
(KDA) window-boundary state re-quantization kernels. Both use CuTe DSL and
|