abetlen / abetlen/llama-cpp-python
Have you thought about adding quantum cache or 8 bit cache?
未關閉
enhancement
- 主要語言
- Python
- 星號
- 10.6k
- 分支
- 1.4k
- PR 合併指標
- PR 指標待擷取
描述
I've figured out how to change the type. It's a little slower but it does save memory.
The numbers for type_K or V simply correspond to the cache types in ggml.h
```
GGML_TYPE_F32 = 0,
GGML_TYPE_F16 = 1,
GGML_TYPE_Q4_0 = 2,
GGML_TYPE_Q4_1 = 3,
// GGML_TYPE_Q4_2 = 4, support has been removed
// GGML_TYPE_Q4_3 (5) support has been removed
GGML_TYPE_Q5_0 = 6,
GGML_TYPE_Q5_1 = 7,
GGML_TYPE_Q8_0 = 8,
GGML_TYPE_Q8_1 = 9,
```
貢獻指南
評估
這個 Issue 還沒有評估資料。