Update README.md
Browse files
README.md
CHANGED
|
@@ -1,5 +1,10 @@
|
|
| 1 |
---
|
| 2 |
license: mit
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
| 4 |
|
| 5 |
EXL3 quants of [MiniMax-M2](https://huggingface.co/MiniMaxAI/MiniMax-M2)
|
|
@@ -8,10 +13,29 @@ EXL3 quants of [MiniMax-M2](https://huggingface.co/MiniMaxAI/MiniMax-M2)
|
|
| 8 |
|
| 9 |
Base bitrates:
|
| 10 |
|
|
|
|
| 11 |
[3.00 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/3.0bpw)
|
| 12 |
[4.00 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/4.0bpw)
|
| 13 |
-
(more to come)
|
| 14 |
|
| 15 |
Optimized:
|
| 16 |
|
| 17 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: mit
|
| 3 |
+
base_model: MiniMaxAI/MiniMax-M2
|
| 4 |
+
base_model_relation: quantized
|
| 5 |
+
quantized_by: turboderp
|
| 6 |
+
tags:
|
| 7 |
+
- exl3
|
| 8 |
---
|
| 9 |
|
| 10 |
EXL3 quants of [MiniMax-M2](https://huggingface.co/MiniMaxAI/MiniMax-M2)
|
|
|
|
| 13 |
|
| 14 |
Base bitrates:
|
| 15 |
|
| 16 |
+
[2.00 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/2.0bpw)
|
| 17 |
[3.00 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/3.0bpw)
|
| 18 |
[4.00 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/4.0bpw)
|
|
|
|
| 19 |
|
| 20 |
Optimized:
|
| 21 |
|
| 22 |
+
[2.04 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/2.04bpw)
|
| 23 |
+
[2.27 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/2.27bpw)
|
| 24 |
+
[3.04 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/3.04bpw)
|
| 25 |
+
[3.50 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/3.5bpw)
|
| 26 |
+
[4.03 bits per weight](https://huggingface.co/turboderp/MiniMax-M2-exl3/tree/4.03bpw)
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
. | KL-div | ppl | HumanEval@1
|
| 30 |
+
---------|--------|-------|-------------
|
| 31 |
+
2.00 bpw | 0.400 | 10.92 | 80.5%
|
| 32 |
+
2.04 bpw | 0.297 | 10.23 | 87.1%
|
| 33 |
+
2.27 bpw | 0.252 | 9.78 | 88.4%
|
| 34 |
+
3.00 bpw | 0.141 | 8.99 | 87.8%
|
| 35 |
+
3.04 bpw | 0.117 | 8.73 | 87.2%
|
| 36 |
+
3.50 bpw | 0.094 | 8.78 | 88.4%
|
| 37 |
+
4.00 bpw | 0.087 | 8.58 | 89.6%
|
| 38 |
+
4.03 bpw | 0.077 | 8.61 |
|
| 39 |
+
original | - | 8.51 | 87.2%¹
|
| 40 |
+
|
| 41 |
+
¹ Unconfirmed
|