Update README.md
Browse files
README.md
CHANGED
@@ -17,18 +17,7 @@ Dataset Entry Rows:
|
|
17 |
<br>General: 8.7K
|
18 |
<br>Total: 28.15K
|
19 |
|
20 |
-
This is after heavy filtering of ~500K Rows and Entries
|
21 |
-
<br> V2 will see this further reduced down to ~10K after I do a second round of cleaning.
|
22 |
-
|
23 |
-
Applicable Formats:
|
24 |
-
|
25 |
-
ShareGPT / Vicuna
|
26 |
-
<br>Alpaca
|
27 |
-
|
28 |
-
V1 is trained on 50/50 for these two formats.
|
29 |
-
<br>I am working on converting to either for v2.
|
30 |
-
|
31 |
-
Once V2 is Completed, I will also train a 70B variant of this.
|
32 |
|
33 |
v1 is simply an experimental release. V2 will be the main product?
|
34 |
<br>Goals:
|
@@ -41,26 +30,23 @@ v1 is simply an experimental release. V2 will be the main product?
|
|
41 |
|
42 |
I recommend keeping Repetition Penalty below 1.1, preferably at 1 as Hesperus begins breaking down at 1.2 Rep Pen and might output nonsense outputs.
|
43 |
|
|
|
|
|
44 |
Prompt Format:
|
45 |
```
|
46 |
-
- sharegpt
|
47 |
User:
|
48 |
-
GPT:
|
49 |
```
|
50 |
```
|
51 |
-
- alpaca
|
52 |
###Instruction:
|
53 |
Your instruction or question here.
|
54 |
For roleplay purposes, I suggest the following - Write <CHAR NAME>'s next reply in a chat between <YOUR NAME> and <CHAR NAME>. Write a single reply only.
|
55 |
###Response:
|
56 |
```
|
57 |
|
58 |
-
|
59 |
-
|
60 |
-
This model was trained from scratch on the Hesperus dataset.
|
61 |
-
It achieves the following results on the evaluation set:
|
62 |
-
- Loss: 1.5134
|
63 |
-
|
64 |
-
***
|
65 |
|
66 |
-
|
|
|
17 |
<br>General: 8.7K
|
18 |
<br>Total: 28.15K
|
19 |
|
20 |
+
This is after heavy filtering of ~500K Rows and Entries from randomly selected scraped sites and datasets.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
v1 is simply an experimental release. V2 will be the main product?
|
23 |
<br>Goals:
|
|
|
30 |
|
31 |
I recommend keeping Repetition Penalty below 1.1, preferably at 1 as Hesperus begins breaking down at 1.2 Rep Pen and might output nonsense outputs.
|
32 |
|
33 |
+
![Format](https://i.gyazo.com/b22ba269e509c8a62276cbd5bde5acef.png)
|
34 |
+
|
35 |
Prompt Format:
|
36 |
```
|
37 |
+
- sharegpt (recommended!)
|
38 |
User:
|
39 |
+
GPT:
|
40 |
```
|
41 |
```
|
42 |
+
- alpaca (less recommended)
|
43 |
###Instruction:
|
44 |
Your instruction or question here.
|
45 |
For roleplay purposes, I suggest the following - Write <CHAR NAME>'s next reply in a chat between <YOUR NAME> and <CHAR NAME>. Write a single reply only.
|
46 |
###Response:
|
47 |
```
|
48 |
|
49 |
+
V1 is trained on 50/50 for these two formats.
|
50 |
+
<br>I am working on converting to either for v2.
|
|
|
|
|
|
|
|
|
|
|
51 |
|
52 |
+
Once V2 is Completed, I will also train a 70B variant of this.
|