mirror of
https://github.com/ikawrakow/ik_llama.cpp.git
synced 2026-05-01 11:51:53 +00:00
15 lines
373 B
Markdown
15 lines
373 B
Markdown
### 🔀 [#190](https://github.com/ikawrakow/ik_llama.cpp/pull/190) - cuda: non-contiguous rms norm
|
|
|
|
| **Author** | `ikawrakow` |
|
|
| :--- | :--- |
|
|
| **State** | ❌ **Closed** |
|
|
| **Created** | 2025-02-06 |
|
|
| **Updated** | 2025-02-07 |
|
|
|
|
---
|
|
|
|
#### Description
|
|
|
|
Derived from https://github.com/ggerganov/llama.cpp/pull/11659
|
|
|
|
Minor benefit for DeepSeek-Lite (~2% faster TG). |