Update README.md
Browse files
README.md
CHANGED
|
@@ -6,8 +6,6 @@ tags:
|
|
| 6 |
- ik_llama.cpp
|
| 7 |
---
|
| 8 |
|
| 9 |
-
Unfortunately there are some issues with tokenizer and I am quite sure it is because of imatrix in the bf16 repo. I tried using the model and it is coherent but I have no idea if it affects the quality. I will probably try to make an imatrix myself later on and requant it.
|
| 10 |
-
|
| 11 |
IQ2_KS quant of DeepSeek-R1 I made for my 192GB DDR5 + 3090/4090. Done according to:
|
| 12 |
|
| 13 |
#### `IQ2_KS` 183.004 GiB (2.339 BPW)
|
|
|
|
| 6 |
- ik_llama.cpp
|
| 7 |
---
|
| 8 |
|
|
|
|
|
|
|
| 9 |
IQ2_KS quant of DeepSeek-R1 I made for my 192GB DDR5 + 3090/4090. Done according to:
|
| 10 |
|
| 11 |
#### `IQ2_KS` 183.004 GiB (2.339 BPW)
|