mirror of
https://github.com/ikawrakow/ik_llama.cpp.git
synced 2026-01-26 17:20:01 +00:00
17 lines
464 B
Markdown
17 lines
464 B
Markdown
### 🔀 [#475](https://github.com/ikawrakow/ik_llama.cpp/pull/475) - Metal implementatio for the trellis quants.
|
|
|
|
| **Author** | `ikawrakow` |
|
|
| :--- | :--- |
|
|
| **State** | ❌ **Closed** |
|
|
| **Created** | 2025-05-30 |
|
|
| **Updated** | 2025-06-01 |
|
|
|
|
---
|
|
|
|
#### Description
|
|
|
|
`IQ2_KT` and `IQ3_KT` work. `IQ2_KT` has a pretty decent performance.
|
|
|
|
~`IQ4_KT` is not working, so a draft PR for now.~
|
|
|
|
`IQ4_KT` is disabled for now as there is a bug that I don't find. |