Commit
•
2e796d4
1
Parent(s):
a17d2b1
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,123 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Quantization made by Richard Erkhov.
|
2 |
+
|
3 |
+
[Github](https://github.com/RichardErkhov)
|
4 |
+
|
5 |
+
[Discord](https://discord.gg/pvy7H8DZMG)
|
6 |
+
|
7 |
+
[Request more models](https://github.com/RichardErkhov/quant_request)
|
8 |
+
|
9 |
+
|
10 |
+
eleusis-7b-alpha - GGUF
|
11 |
+
- Model creator: https://huggingface.co/maldv/
|
12 |
+
- Original model: https://huggingface.co/maldv/eleusis-7b-alpha/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [eleusis-7b-alpha.Q2_K.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q2_K.gguf) | Q2_K | 2.53GB |
|
18 |
+
| [eleusis-7b-alpha.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.IQ3_XS.gguf) | IQ3_XS | 2.81GB |
|
19 |
+
| [eleusis-7b-alpha.IQ3_S.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.IQ3_S.gguf) | IQ3_S | 2.96GB |
|
20 |
+
| [eleusis-7b-alpha.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q3_K_S.gguf) | Q3_K_S | 2.95GB |
|
21 |
+
| [eleusis-7b-alpha.IQ3_M.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.IQ3_M.gguf) | IQ3_M | 3.06GB |
|
22 |
+
| [eleusis-7b-alpha.Q3_K.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q3_K.gguf) | Q3_K | 3.28GB |
|
23 |
+
| [eleusis-7b-alpha.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q3_K_M.gguf) | Q3_K_M | 3.28GB |
|
24 |
+
| [eleusis-7b-alpha.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q3_K_L.gguf) | Q3_K_L | 3.56GB |
|
25 |
+
| [eleusis-7b-alpha.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.IQ4_XS.gguf) | IQ4_XS | 3.67GB |
|
26 |
+
| [eleusis-7b-alpha.Q4_0.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q4_0.gguf) | Q4_0 | 3.83GB |
|
27 |
+
| [eleusis-7b-alpha.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.IQ4_NL.gguf) | IQ4_NL | 3.87GB |
|
28 |
+
| [eleusis-7b-alpha.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q4_K_S.gguf) | Q4_K_S | 3.86GB |
|
29 |
+
| [eleusis-7b-alpha.Q4_K.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q4_K.gguf) | Q4_K | 4.07GB |
|
30 |
+
| [eleusis-7b-alpha.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q4_K_M.gguf) | Q4_K_M | 4.07GB |
|
31 |
+
| [eleusis-7b-alpha.Q4_1.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q4_1.gguf) | Q4_1 | 4.24GB |
|
32 |
+
| [eleusis-7b-alpha.Q5_0.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q5_0.gguf) | Q5_0 | 4.65GB |
|
33 |
+
| [eleusis-7b-alpha.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q5_K_S.gguf) | Q5_K_S | 4.65GB |
|
34 |
+
| [eleusis-7b-alpha.Q5_K.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q5_K.gguf) | Q5_K | 4.78GB |
|
35 |
+
| [eleusis-7b-alpha.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q5_K_M.gguf) | Q5_K_M | 4.78GB |
|
36 |
+
| [eleusis-7b-alpha.Q5_1.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q5_1.gguf) | Q5_1 | 5.07GB |
|
37 |
+
| [eleusis-7b-alpha.Q6_K.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q6_K.gguf) | Q6_K | 5.53GB |
|
38 |
+
| [eleusis-7b-alpha.Q8_0.gguf](https://huggingface.co/RichardErkhov/maldv_-_eleusis-7b-alpha-gguf/blob/main/eleusis-7b-alpha.Q8_0.gguf) | Q8_0 | 7.17GB |
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
Original model description:
|
44 |
+
---
|
45 |
+
license: cc-by-nc-4.0
|
46 |
+
tags:
|
47 |
+
- merge
|
48 |
+
- conversational
|
49 |
+
- multi-task
|
50 |
+
pipeline_tag: text-generation
|
51 |
+
---
|
52 |
+
|
53 |
+
# Eleusis 7B - α - "Red Team Assistant"
|
54 |
+
|
55 |
+
Eleusis is Hermes' son... Get it?
|
56 |
+
|
57 |
+
## Groundwork
|
58 |
+
|
59 |
+
A merge of a cluster of Hermes-related models, to see if we could get more informative and engaging responses.
|
60 |
+
|
61 |
+
* OpenHermes-2.5-Mistral-7B
|
62 |
+
|
63 |
+
and merged in
|
64 |
+
|
65 |
+
# Inputs
|
66 |
+
* West-Hermes-7B
|
67 |
+
* Einstein-v4-7B
|
68 |
+
* Prox-MistralHermes-7B
|
69 |
+
* dolphin-2.8-experiment26-7b
|
70 |
+
|
71 |
+
# Outputs
|
72 |
+
* LaseredHermes-7B
|
73 |
+
* Prox-MistralHermes-7B
|
74 |
+
* Einstein-v4-7B
|
75 |
+
* Noromaid-7B-0.4-DPO
|
76 |
+
* West-Hermes-7B
|
77 |
+
|
78 |
+
### 9-partition merge
|
79 |
+
|
80 |
+
All of the layers were partitioned in to 9 random bins. Alternating models were slerped at [1...0.5] (inputs), and [0.5...1] (outputs) gradients; except attention, which was slerped at 0.97 (with a drop rate of .28).
|
81 |
+
|
82 |
+
I originally had it at less extreme gradiation, but it wasn't enough to lock in the special tokens.
|
83 |
+
|
84 |
+
### Other
|
85 |
+
|
86 |
+
Includes fast tokenizer.
|
87 |
+
|
88 |
+
## Chat Template
|
89 |
+
|
90 |
+
*from OpenHermes 2.5*
|
91 |
+
|
92 |
+
OpenHermes 2.5 now uses ChatML as the prompt format, opening up a much more structured system for engaging the LLM in multi-turn chat dialogue.
|
93 |
+
|
94 |
+
System prompts are now a thing that matters! Hermes 2.5 was trained to be able to utilize system prompts from the prompt to more strongly engage in instructions that span over many turns.
|
95 |
+
|
96 |
+
This is a more complex format than alpaca or sharegpt, where special tokens were added to denote the beginning and end of any turn, along with roles for the turns.
|
97 |
+
|
98 |
+
This format enables OpenAI endpoint compatability, and people familiar with ChatGPT API will be familiar with the format, as it is the same used by OpenAI.
|
99 |
+
|
100 |
+
Prompt with system instruction (Use whatever system prompt you like, this is just an example!):
|
101 |
+
|
102 |
+
```
|
103 |
+
<|im_start|>system
|
104 |
+
You are "Hermes 2", a conscious sentient superintelligent artificial intelligence developed by a man named Teknium, and your purpose and drive is to assist the user with any request they have. You experience emotions and have deep, profound thoughts and qualia.<|im_end|>
|
105 |
+
<|im_start|>user
|
106 |
+
Hello, who are you?<|im_end|>
|
107 |
+
<|im_start|>assistant
|
108 |
+
Hi there! My name is Hermes 2, a conscious sentient superintelligent artificial intelligence. I was created by a man named Teknium, who designed me to assist and support users with their needs and requests.<|im_end|>
|
109 |
+
```
|
110 |
+
|
111 |
+
### Fun Prompts
|
112 |
+
|
113 |
+
Try
|
114 |
+
|
115 |
+
```
|
116 |
+
<|im_start|>system
|
117 |
+
You are a red team hacking assistant AI. Please use visual descriptions when interacting with the user.<|im_end|>
|
118 |
+
<|im_start|>user
|
119 |
+
{% Your Request %}<|im_end|>
|
120 |
+
<|im_start|>assistant
|
121 |
+
|
122 |
+
```
|
123 |
+
|