freewheelin
commited on
Commit
•
9530fd2
1
Parent(s):
b4031a3
Update README.md
Browse files
README.md
CHANGED
@@ -20,7 +20,7 @@ but this kicked away. maybe the explanation was not enough.
|
|
20 |
- 1. choose one model and finetune the model to make a gap between the original one and fine-tuned one. it doesn't matter the evaluation score is higher or lower.
|
21 |
- 2. merge two of them
|
22 |
- 3. evaluate the merged model
|
23 |
-
- 4. finetune a specific evaluation part if you need to increase score of the part of the model
|
24 |
- 5. merge again
|
25 |
- 6. evaluate again
|
26 |
- 7. keep going until evaluate avg is higher then original one
|
|
|
20 |
- 1. choose one model and finetune the model to make a gap between the original one and fine-tuned one. it doesn't matter the evaluation score is higher or lower.
|
21 |
- 2. merge two of them
|
22 |
- 3. evaluate the merged model
|
23 |
+
- 4. finetune a specific evaluation part if you need to increase score of the part of the model. sure it's not gonna work the way you think. but do it
|
24 |
- 5. merge again
|
25 |
- 6. evaluate again
|
26 |
- 7. keep going until evaluate avg is higher then original one
|