Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,44 @@
|
|
1 |
---
|
2 |
license: llama2
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: llama2
|
3 |
---
|
4 |
+
---
|
5 |
+
license: llama2
|
6 |
+
---
|
7 |
+
This is the GGML version of LaMA 2 Holomax 13B for use with [Koboldcpp](https://koboldai.org/cpp) for the original HF compatible release please check [KoboldAI/LLaMA2-13B-Holomax](https://huggingface.co/KoboldAI/LLaMA2-13B-Holomax)
|
8 |
+
|
9 |
+
# LLaMA 2 Holomax 13B - The writers version of Mythomax
|
10 |
+
|
11 |
+
This is an expansion merge to the well praised Mythomax model from Gryphe (60%) using MrSeeker's KoboldAI Holodeck model (40%)
|
12 |
+
The goal of this model is to enhance story writing capabilities while preserving the desirable traits of the Mythomax model.
|
13 |
+
|
14 |
+
Testers found that this model passes the InteracTV benchmark, was useful for story writing, chatting and text adventures using Instruction mode.
|
15 |
+
Preservation of factual knowledge has not been tested since we expect the original to be better in those use cases as this merge was focussed on fiction.
|
16 |
+
|
17 |
+
## Credits
|
18 |
+
This merge is not possible without the following models and model authors (Thanks to all of you for your work!)
|
19 |
+
|
20 |
+
Mythomax by Gryphe:
|
21 |
+
- Mythologic-L2 by Gryphe:
|
22 |
+
- - Hermes by Nous-Research
|
23 |
+
- Chronos V2 by Elinas
|
24 |
+
- Airoboros m2.0 by Jondurbin
|
25 |
+
- Huginn by Face of Goonery:
|
26 |
+
- - Hermes by Nous-Research
|
27 |
+
- StableBeluga by StabilityAI
|
28 |
+
- Airoboros by Jondurbin
|
29 |
+
- Chronos by Elinas
|
30 |
+
- Limarp by Lemonila
|
31 |
+
|
32 |
+
Holodeck by Mr.Seeker
|
33 |
+
|
34 |
+
## Guidelines
|
35 |
+
This model is designed to be flexible, it should be able to be used as a co-writing model, as well as a variety of instruct formats (Tested with Alpaca) and regular chatting both augmented with traditional formatting and instruct formatting.
|
36 |
+
The Alpaca format is as follows:
|
37 |
+
```
|
38 |
+
### Instruction:
|
39 |
+
|
40 |
+
Instruction goes here
|
41 |
+
|
42 |
+
### Response:
|
43 |
+
```
|
44 |
+
But if you have a different preferred format that works on one of the models above it will likely still work.
|