Commit
•
43bd337
1
Parent(s):
4b29762
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,75 @@
|
|
1 |
---
|
2 |
license: cc-by-nc-4.0
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: cc-by-nc-4.0
|
3 |
+
tags:
|
4 |
+
- GGUF
|
5 |
+
- iMat
|
6 |
+
- llama3
|
7 |
---
|
8 |
+
|
9 |
+
```
|
10 |
+
e88 88e d8
|
11 |
+
d888 888b 8888 8888 ,"Y88b 888 8e d88
|
12 |
+
C8888 8888D 8888 8888 "8" 888 888 88b d88888
|
13 |
+
Y888 888P Y888 888P ,ee 888 888 888 888
|
14 |
+
"88 88" "88 88" "88 888 888 888 888
|
15 |
+
b
|
16 |
+
8b,
|
17 |
+
|
18 |
+
e88'Y88 d8 888
|
19 |
+
d888 'Y ,"Y88b 888,8, d88 ,e e, 888
|
20 |
+
C8888 "8" 888 888 " d88888 d88 88b 888
|
21 |
+
Y888 ,d ,ee 888 888 888 888 , 888
|
22 |
+
"88,d88 "88 888 888 888 "YeeP" 888
|
23 |
+
|
24 |
+
PROUDLY PRESENTS
|
25 |
+
```
|
26 |
+
|
27 |
+
## Llama-3-8B-EGO-iMat-GGUF
|
28 |
+
|
29 |
+
|
30 |
+
Quantized from fp32 with love.
|
31 |
+
* Weighted quantizations were calculated using groups_merged.txt with 105 chunks (recommended amount for this file) and n_ctx=512. Special thanks to jukofyork for sharing [this process](https://huggingface.co/jukofyork/WizardLM-2-8x22B-imatrix)
|
32 |
+
|
33 |
+
<b>**Note - Please use SillyTavern as well as the following prompt format:**</b>
|
34 |
+
```
|
35 |
+
[EGO]Name: Character name and then Everything that forms the personality and speech patterns.(i.e. scenario, sample dialogue, character definitions, etc)[/EGO]
|
36 |
+
[SEEN]User message.[/SEEN]
|
37 |
+
Character Name:
|
38 |
+
```
|
39 |
+
|
40 |
+
|
41 |
+
For a brief rundown of iMatrix quant performance please see this [PR](https://github.com/ggerganov/llama.cpp/pull/5747)
|
42 |
+
|
43 |
+
<b>All quants are verified working prior to uploading to repo for your safety and convenience. </b>
|
44 |
+
|
45 |
+
Please note importance matrix quantizations are a work in progress. IQ4 and above is recommended for best results.
|
46 |
+
|
47 |
+
Original model card [here](https://huggingface.co/Envoid/Llama-3-8B-EGO) and below
|
48 |
+
|
49 |
+
---
|
50 |
+
|
51 |
+
# This model isn't particularly great. It's just an undercooked experiment.
|
52 |
+
|
53 |
+
Releasing it anyways just in case it accidentally makes good merge meat.
|
54 |
+
|
55 |
+
# It also has a tendency to produce mature content without warning.
|
56 |
+
|
57 |
+
This model is tuned off of the base Llama-3-8B model.
|
58 |
+
|
59 |
+
I adapted the leaked Undi dataset into training samples for custom formatting. This model pretty much only functions properly in SillyTavern.
|
60 |
+
|
61 |
+
The formatting has two pairs of pseudotokens
|
62 |
+
|
63 |
+
```
|
64 |
+
[EGO]Name: Character name and then Everything that forms the personality and speech patterns.(i.e. scenario, sample dialogue, character definitions, etc)[/EGO]
|
65 |
+
[SEEN]User message.[/SEEN]
|
66 |
+
Character Name:
|
67 |
+
```
|
68 |
+
|
69 |
+
The self attention modules were fine tuned separately on this dataset and the pseudotokens were chosen because they made logical sense with respect to the character giving a reply without allowing the model to 'connect the dots' during training and figure out that it is indeed an AI language model.
|
70 |
+
|
71 |
+
After this was done all modules were then finetuned together on the dendrite dataset in order to connect the changes made to the attention modules.
|
72 |
+
|
73 |
+
So with regards to building a SillyTavern prompt template you basically want the entire story string and any additional stylistic instructions enclosed in the [EGO] tags and then the user messages enclosed in [SEEN] tags.
|
74 |
+
|
75 |
+
It doesn't give particularly verbose replies unless you're continueing a roleplay with verbose messages. Otherwise it's pretty bad.
|