Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
tags:
|
3 |
+
- llama
|
4 |
+
- alpaca
|
5 |
+
- vicuna
|
6 |
+
- uncensored
|
7 |
+
- cot
|
8 |
+
- chain of thought
|
9 |
+
- story
|
10 |
+
- adventure
|
11 |
+
- roleplay
|
12 |
+
- rp
|
13 |
+
- merge
|
14 |
+
- mix
|
15 |
+
- instruct
|
16 |
+
- wizardlm
|
17 |
+
- superhot
|
18 |
+
- supercot
|
19 |
+
- manticore
|
20 |
+
- hippogriff
|
21 |
+
---
|
22 |
+
|
23 |
+
## 30B-Epsilon
|
24 |
+
|
25 |
+
Epsilon is an instruct based general purpose model assembled from hand picked models and LoRAs.
|
26 |
+
There is no censorship and it follows instructions in the Alpaca format. This means you can create
|
27 |
+
your own rules in the context memory of your inference system of choice [mainly KoboldAI or Text
|
28 |
+
Generation Webui and chat UIs like SillyTavern and so on].
|
29 |
+
|
30 |
+
## Composition:
|
31 |
+
|
32 |
+
This model is the result of an experimental use of LoRAs on language models and model merges.
|
33 |
+
[] = applied as LoRA to a composite model | () = combined as composite models
|
34 |
+
30B-Epsilon = [SuperCOT[SuperHOT-prototype13b-8192[(wizardlmuncensored+((hippogriff+manticore)+(StoryV2))]
|
35 |
+
|
36 |
+
Alpaca's instruct format can be used to do many things, including control of the terms of behavior
|
37 |
+
between a user and a response from an agent in chat. Below is an example of a command injected into
|
38 |
+
memory.
|
39 |
+
|
40 |
+
```
|
41 |
+
### Instruction:
|
42 |
+
Make Narrator function as a text based adventure game that responds with verbose, detailed, and creative descriptions of what happens next after Player's response.
|
43 |
+
Make Player function as the player input for Narrator's text based adventure game, controlling a character named (insert character name here, their short bio, and
|
44 |
+
whatever quest or other information to keep consistent in the interaction).
|
45 |
+
|
46 |
+
### Response:
|
47 |
+
{an empty new line here}
|
48 |
+
```
|
49 |
+
|
50 |
+
All datasets from all models and LoRAs used were documented and reviewed as model candidates for merging.
|
51 |
+
Model candidates were based on five core principles: creativity, logic, inference, instruction following,
|
52 |
+
and longevity of trained responses. SuperHOTv2 did not have datasets public at the time it was selected,
|
53 |
+
however it was included as a key component of Epsilon considering the datasets its predecessor
|
54 |
+
included, which are no longer available to view as it has been removed [kaiokendev-SuperHOT-LoRA-prototype30b-8192].
|
55 |
+
SuperHot 2 is utilized regardless in light of that. The GPT4Alpaca LoRA from Chansung was removed from this
|
56 |
+
amalgam following a thorough review of where censorship and railroading the user came from in 33B-Lazarus.
|
57 |
+
This is not a reflection of ChanSung's excellent work - it merely did not fit the purpose of this model.
|
58 |
+
|
59 |
+
## Language Models and LoRAs Used Credits:
|
60 |
+
|
61 |
+
manticore-30b-chat-pyg-alpha [Epoch0.4] by openaccess-ai-collective
|
62 |
+
|
63 |
+
https://huggingface.co/openaccess-ai-collective/manticore-30b-chat-pyg-alpha
|
64 |
+
|
65 |
+
hippogriff-30b-chat by openaccess-ai-collective
|
66 |
+
|
67 |
+
https://huggingface.co/openaccess-ai-collective/hippogriff-30b-chat
|
68 |
+
|
69 |
+
WizardLM-33B-V1.0-Uncensored by ehartford
|
70 |
+
|
71 |
+
https://huggingface.co/ehartford/WizardLM-33B-V1.0-Uncensored
|
72 |
+
|
73 |
+
Storytelling-LLaMa-LoRA [30B, Version 2] by GamerUnTouch
|
74 |
+
|
75 |
+
https://huggingface.co/GamerUntouch/Storytelling-LLaMa-LoRAs
|
76 |
+
|
77 |
+
SuperCOT-LoRA [30B] by kaiokendev
|
78 |
+
|
79 |
+
https://huggingface.co/kaiokendev/SuperCOT-LoRA
|
80 |
+
|
81 |
+
SuperHOT-LoRA-prototype30b-8192 [30b, not 8K version, but a removed prototype] by kaiokendev
|
82 |
+
|
83 |
+
https://huggingface.co/kaiokendev/superhot-30b-8k-no-rlhf-test [Similar LoRA to one since removed that was used in making this model.]
|
84 |
+
|
85 |
+
Also thanks to Meta for LLaMA and to each and every one of you
|
86 |
+
who developed these fine-tunes and LoRAs.
|