Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model:
|
| 3 |
+
- MiniMaxAI/MiniMax-M2.1
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
This is a custom quant of [MiniMaxAI/MiniMax-M2.1](https://huggingface.co/MiniMaxAI/MiniMax-M2.1) that has the following:
|
| 7 |
+
- Q8_0 for the default quantization type (attention, shared experts, etc.)
|
| 8 |
+
- Q4_K for the FFN_UP and FFN_GATE tensors
|
| 9 |
+
- Q5_K for the FFN_DOWN tensors
|
| 10 |
+
|
| 11 |
+
The idea being that given the huge size of the FFN tensors compared to the rest of the tensors in the model, it should be possible to achieve a better quality while keeping the overall size of the entire model smaller compared to a similar naive quantization.
|
| 12 |
+
|
| 13 |
+
This model was produced using [Bartowski's imatrix](https://huggingface.co/bartowski/MiniMaxAI_MiniMax-M2.1-GGUF/blob/main/MiniMaxAI_MiniMax-M2.1-imatrix.gguf).
|
| 14 |
+
|
| 15 |
+
Model is additionally split with `--no-tensor-first-split` to enable easier editing of metadata.
|