Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,95 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-nc-4.0
|
3 |
+
---
|
4 |
+
## Exl2 version of [NeverSleep/Noromaid-v0.4-Mixtral-Instruct-8x7b-Zloss](https://huggingface.co/NeverSleep/Noromaid-v0.4-Mixtral-Instruct-8x7b-Zloss)
|
5 |
+
|
6 |
+
## branch
|
7 |
+
3.5bh8 : 3.5bpw h8
|
8 |
+
|
9 |
+
Using ThePile [0007.parquet](https://huggingface.co/datasets/EleutherAI/the_pile_deduplicated/resolve/refs%2Fconvert%2Fparquet/default/train/0007.parquet) as dataset
|
10 |
+
|
11 |
+
Quantization settings : ```python convert.py -i models/NeverSleep_Noromaid-v0.4-Mixtral-Instruct-8x7b-Zloss -o Noromaid-v0.4-Mixtral-Instruct-8x7b-Zloss-temp4 -cf Noromaid-v0.4-Mixtral-Instruct-8x7b-Zloss-3.5bpw-h8-exl2 -c 0007.parquet -l 8192 -b 3.5 -hb 8 -ml 8192```
|
12 |
+
### below this line is original readme
|
13 |
+
|
14 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/630dfb008df86f1e5becadc3/vwcJfOnL-2QDJ0ShfxRJ5.png)
|
15 |
+
|
16 |
+
|
17 |
+
|
18 |
+
---
|
19 |
+
|
20 |
+
# Disclaimer:
|
21 |
+
## This model is experimental, do not expect everything to work.
|
22 |
+
|
23 |
+
This model uses the Chatml **prompting format**
|
24 |
+
|
25 |
+
---
|
26 |
+
|
27 |
+
|
28 |
+
Beeg noromaid on ***steroids***. Suitable for RP, ERP.
|
29 |
+
|
30 |
+
This model was trained on the Zloss fork of Charles, and should fix issue the model had.
|
31 |
+
|
32 |
+
Use Chatml prompt format, but not the special token.
|
33 |
+
|
34 |
+
The reason is that Axolotl merge the finetune with the base model at 1.0 weight basically, but this is too much, so I use another script available [HERE](https://github.com/DocShotgun/LLM-notebooks/blob/main/weighted-lora-merge.ipynb) to merge with less weight, sadly, it don't take the special Chatml token. It's like Orca2 for the matter.
|
35 |
+
|
36 |
+
|
37 |
+
## Credits:
|
38 |
+
- Undi
|
39 |
+
- IkariDev
|
40 |
+
|
41 |
+
<!-- description start -->
|
42 |
+
## Description
|
43 |
+
|
44 |
+
<!-- [Recommended settings - contributed by localfultonextractor](https://files.catbox.moe/ue0tja.json) -->
|
45 |
+
|
46 |
+
This repo contains FP16 files of Noromaid-v0.4-Mixtral-Instruct-8x7b-Zloss.
|
47 |
+
|
48 |
+
[FP16 - by IkariDev and Undi](https://huggingface.co/NeverSleep/Noromaid-v0.4-Mixtral-Instruct-8x7b-Zloss)
|
49 |
+
|
50 |
+
<!-- [GGUF - By TheBloke](https://huggingface.co/TheBloke/Athena-v4-GGUF)-->
|
51 |
+
|
52 |
+
<!-- [GPTQ - By TheBloke](https://huggingface.co/TheBloke/Athena-v4-GPTQ)-->
|
53 |
+
|
54 |
+
<!-- [exl2[8bpw-8h] - by AzureBlack](https://huggingface.co/AzureBlack/Echidna-13b-v0.3-8bpw-8h-exl2)-->
|
55 |
+
|
56 |
+
<!-- [AWQ - By TheBloke](https://huggingface.co/TheBloke/Athena-v4-AWQ)-->
|
57 |
+
|
58 |
+
<!-- [fp16 - by IkariDev+Undi95](https://huggingface.co/IkariDev/Athena-v4)-->
|
59 |
+
|
60 |
+
[GGUF - by IkariDev and Undi](https://huggingface.co/NeverSleep/Noromaid-v0.4-Mixtral-Instruct-8x7b-Zloss-GGUF)
|
61 |
+
<!-- [OLD(GGUF - by IkariDev+Undi95)](https://huggingface.co/IkariDev/Athena-v4-GGUF)-->
|
62 |
+
|
63 |
+
## Ratings:
|
64 |
+
|
65 |
+
Note: We have permission of all users to upload their ratings, we DONT screenshot random reviews without asking if we can put them here!
|
66 |
+
|
67 |
+
No ratings yet!
|
68 |
+
|
69 |
+
If you want your rating to be here, send us a message over on DC and we'll put up a screenshot of it here. DC name is "ikaridev" and "undi".
|
70 |
+
|
71 |
+
<!-- description end -->
|
72 |
+
<!-- prompt-template start -->
|
73 |
+
### Prompt format: Chatml
|
74 |
+
```
|
75 |
+
<|im_start|>system
|
76 |
+
{sysprompt}<|im_end|>
|
77 |
+
<|im_start|>user
|
78 |
+
{input}<|im_end|>
|
79 |
+
<|im_start|>assistant
|
80 |
+
{output}<|im_end|>
|
81 |
+
```
|
82 |
+
|
83 |
+
## Datasets used:
|
84 |
+
|
85 |
+
- Aesir 1, 2 & 3 modified by us, credit to ([MinervaAI](https://huggingface.co/MinervaAI) / [Gryphe](https://huggingface.co/Gryphe))
|
86 |
+
- [LimaRP-20231109](https://huggingface.co/datasets/lemonilia/LimaRP) ([Lemonilia](https://huggingface.co/lemonilia))
|
87 |
+
- [ToxicQAFinal](https://huggingface.co/datasets/NobodyExistsOnTheInternet/ToxicQAFinal) ([NobodyExistsOnTheInternet](https://huggingface.co/NobodyExistsOnTheInternet)
|
88 |
+
- [No-robots-ShareGPT](https://huggingface.co/datasets/Doctor-Shotgun/no-robots-sharegpt) ([Doctor-Shotgun](https://huggingface.co/Doctor-Shotgun))
|
89 |
+
|
90 |
+
|
91 |
+
## Others
|
92 |
+
|
93 |
+
Undi: If you want to support me, you can [here](https://ko-fi.com/undiai).
|
94 |
+
|
95 |
+
IkariDev: Visit my [retro/neocities style website](https://ikaridevgit.github.io/) please kek
|