freewheelin
commited on
Update README.md
Browse files
README.md
CHANGED
@@ -7,10 +7,23 @@ license: mit
|
|
7 |
|
8 |
# Model Card for free-evo-qwen72b-v0.8
|
9 |
|
10 |
-
##
|
|
|
|
|
11 |
|
12 |
## Method
|
13 |
- We were inspired by this [Sakana project](https://sakana.ai/evolutionary-model-merge/)
|
14 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
## Base Architecture
|
16 |
-
-
|
|
|
7 |
|
8 |
# Model Card for free-evo-qwen72b-v0.8
|
9 |
|
10 |
+
## 1st place : 2024 4th May - avg. 81.28
|
11 |
+
[Open Llm Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
12 |
+
but this kicked away. maybe the explanation was not enough. i don't care.
|
13 |
|
14 |
## Method
|
15 |
- We were inspired by this [Sakana project](https://sakana.ai/evolutionary-model-merge/)
|
16 |
|
17 |
+
## Process
|
18 |
+
|
19 |
+
#### 1. two models with the same architecture are needed so fine-tune a model to create a gap between the two of them.
|
20 |
+
#### 2. merge original one and fine-tuned one
|
21 |
+
#### 3. evaluate the merged model
|
22 |
+
#### 4. merge again it with original model
|
23 |
+
#### 5. evaluate again
|
24 |
+
#### 6. keep going until evaluate avg is higher then original one
|
25 |
+
|
26 |
+
that's it. simple.
|
27 |
+
|
28 |
## Base Architecture
|
29 |
+
- QWEN2
|