Update README.md
Browse files
README.md
CHANGED
|
@@ -5,7 +5,39 @@ tags:
|
|
| 5 |
- merge
|
| 6 |
|
| 7 |
---
|
| 8 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
|
| 11 |
|
|
|
|
| 5 |
- merge
|
| 6 |
|
| 7 |
---
|
| 8 |
+
# friend-8x7B-hf
|
| 9 |
+
|
| 10 |
+
Several-staged merge this time. Actually decent, from my testings.
|
| 11 |
+
|
| 12 |
+
Use ChatML or Alpaca, both seemed to work though I liked the outputs from ChatML more.
|
| 13 |
+
|
| 14 |
+
temp-output-base:
|
| 15 |
+
```yaml
|
| 16 |
+
models:
|
| 17 |
+
- model: mistralai/Mixtral-8x7B-v0.1+/ai/LLM/tmp/pefts/daybreak-peft/mixtral-8x7b
|
| 18 |
+
parameters:
|
| 19 |
+
weight: 0.65
|
| 20 |
+
- model: mistralai/Mixtral-8x7B-v0.1+retrieval-bar/Mixtral-8x7B-v0.1_case-briefs
|
| 21 |
+
parameters:
|
| 22 |
+
weight: 0.25
|
| 23 |
+
merge_method: task_arithmetic
|
| 24 |
+
base_model: mistralai/Mixtral-8x7B-v0.1
|
| 25 |
+
dtype: float16
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
temp-output-instruct:
|
| 29 |
+
```yaml
|
| 30 |
+
models:
|
| 31 |
+
- model: mistralai/Mixtral-8x7B-Instruct-v0.1+SeanWu25/Mixtral_8x7b_Medicine
|
| 32 |
+
parameters:
|
| 33 |
+
weight: 0.33
|
| 34 |
+
- model: NeverSleep/Noromaid-v0.4-Mixtral-Instruct-8x7b-Zloss
|
| 35 |
+
parameters:
|
| 36 |
+
weight: 0.15
|
| 37 |
+
merge_method: task_arithmetic
|
| 38 |
+
base_model: mistralai/Mixtral-8x7B-Instruct-v0.1
|
| 39 |
+
dtype: float16
|
| 40 |
+
```
|
| 41 |
|
| 42 |
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
|
| 43 |
|