Commit
•
3bdbce5
1
Parent(s):
3170216
Update README.md
Browse files
README.md
CHANGED
|
@@ -62,15 +62,13 @@ min-p: 0.02-0.1
|
|
| 62 |
|
| 63 |
## Evals:
|
| 64 |
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
*
|
| 68 |
-
*
|
| 69 |
-
*
|
| 70 |
-
*
|
| 71 |
-
*
|
| 72 |
-
* Winogrande:
|
| 73 |
-
* GSM8K:
|
| 74 |
|
| 75 |
## Examples:
|
| 76 |
```
|
|
@@ -112,7 +110,6 @@ Umbra-v2-MoE-4x10.7 is a Mixure of Experts (MoE) made with the following models:
|
|
| 112 |
|
| 113 |
```python
|
| 114 |
!pip install -qU transformers bitsandbytes accelerate
|
| 115 |
-
|
| 116 |
from transformers import AutoTokenizer
|
| 117 |
import transformers
|
| 118 |
import torch
|
|
|
|
| 62 |
|
| 63 |
## Evals:
|
| 64 |
|
| 65 |
+
* Avg: 73.59
|
| 66 |
+
* ARC: 69.11
|
| 67 |
+
* HellaSwag: 87.57
|
| 68 |
+
* MMLU: 66.48
|
| 69 |
+
* T-QA: 66.75
|
| 70 |
+
* Winogrande: 83.11
|
| 71 |
+
* GSM8K: 68.69
|
|
|
|
|
|
|
| 72 |
|
| 73 |
## Examples:
|
| 74 |
```
|
|
|
|
| 110 |
|
| 111 |
```python
|
| 112 |
!pip install -qU transformers bitsandbytes accelerate
|
|
|
|
| 113 |
from transformers import AutoTokenizer
|
| 114 |
import transformers
|
| 115 |
import torch
|