SicariusSicariiStuff
commited on
Commit
β’
4e7b705
1
Parent(s):
71977ba
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,122 @@
|
|
1 |
-
---
|
2 |
-
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
license: apache-2.0
|
5 |
+
---
|
6 |
+
|
7 |
+
<div align="center">
|
8 |
+
<b style="font-size: 40px;">Zion_Alpha_Instruction_Tuned_GGUF</b>
|
9 |
+
|
10 |
+
|
11 |
+
</div>
|
12 |
+
|
13 |
+
|
14 |
+
<img src="https://i.imgur.com/e1LEQ18.png" alt="Zion_Alpha_Instruction_Tuned_GGUF" style="width: 50%; min-width: 400px; display: block; margin: auto;">
|
15 |
+
|
16 |
+
|
17 |
+
# Model Details
|
18 |
+
|
19 |
+
Zion_Alpha is the first **REAL** Hebrew model in the world. This version WAS fine tuned for tasks. I did the finetune using SOTA techniques and using my insights from years of underwater basket weaving. If you wanna offer me a job, just add me on Facebook.
|
20 |
+
|
21 |
+
# Future Plans
|
22 |
+
I plan to perform a SLERP merge with one of my other fine-tuned models, which has a bit more knowledge about Israeli topics. Additionally, I might create a larger model using MergeKit, but we'll see how it goes.
|
23 |
+
|
24 |
+
# Looking for Sponsors
|
25 |
+
Since all my work is done on-premises, I am constrained by my current hardware. I would greatly appreciate any support in acquiring an A6000, which would enable me to train significantly larger models much faster.
|
26 |
+
|
27 |
+
# Papers?
|
28 |
+
Maybe. We'll see. No promises here π€
|
29 |
+
|
30 |
+
# Contact Details
|
31 |
+
I'm not great at self-marketing (to say the least) and don't have any social media accounts. If you'd like to reach out to me, you can email me at [email protected]. Please note that this email might receive more messages than I can handle, so I apologize in advance if I can't respond to everyone.
|
32 |
+
|
33 |
+
# Versions and QUANTS
|
34 |
+
- Base model: [FP16](https://huggingface.co/SicariusSicariiStuff/Zion_Alpha)
|
35 |
+
- Instruction tuned: [FP16](https://huggingface.co/SicariusSicariiStuff/Zion_Alpha_Instruction_Tuned) | [GGUF](https://huggingface.co/SicariusSicariiStuff/Zion_Alpha_Instruction_Tuned_GGUF)
|
36 |
+
|
37 |
+
|
38 |
+
# Model architecture
|
39 |
+
Based on Mistral 7B. I didn't even bother to alter the tokenizer.
|
40 |
+
|
41 |
+
# The recommended prompt setting is Debug-deterministic:
|
42 |
+
```
|
43 |
+
temperature: 1
|
44 |
+
top_p: 1
|
45 |
+
top_k: 1
|
46 |
+
typical_p: 1
|
47 |
+
min_p: 1
|
48 |
+
repetition_penalty: 1
|
49 |
+
```
|
50 |
+
|
51 |
+
# The recommended instruction template is Mistral:
|
52 |
+
```
|
53 |
+
{%- for message in messages %}
|
54 |
+
{%- if message['role'] == 'system' -%}
|
55 |
+
{{- message['content'] -}}
|
56 |
+
{%- else -%}
|
57 |
+
{%- if message['role'] == 'user' -%}
|
58 |
+
{{-'[INST] ' + message['content'].rstrip() + ' [/INST]'-}}
|
59 |
+
{%- else -%}
|
60 |
+
{{-'' + message['content'] + '</s>' -}}
|
61 |
+
{%- endif -%}
|
62 |
+
{%- endif -%}
|
63 |
+
{%- endfor -%}
|
64 |
+
{%- if add_generation_prompt -%}
|
65 |
+
{{-''-}}
|
66 |
+
{%- endif -%}
|
67 |
+
```
|
68 |
+
# English to hebrew example:
|
69 |
+
|
70 |
+
|
71 |
+
<div align="center">
|
72 |
+
<b style="font-size: 40px;">Zion_Alpha English to Hebrew example</b>
|
73 |
+
|
74 |
+
|
75 |
+
</div>
|
76 |
+
|
77 |
+
|
78 |
+
<img src="https://i.imgur.com/JnTuawF.png" alt="Zion_Alpha" style="width: 40%; min-width: 600px; display: block; margin: auto;">
|
79 |
+
|
80 |
+
|
81 |
+
# English to hebrew example:
|
82 |
+
|
83 |
+
|
84 |
+
<div align="center">
|
85 |
+
<b style="font-size: 40px;">Zion_Alpha Hebrew to English example</b>
|
86 |
+
|
87 |
+
|
88 |
+
</div>
|
89 |
+
|
90 |
+
|
91 |
+
<img src="https://i.imgur.com/Wm2igLJ.png" alt="Zion_Alpha" style="width: 40%; min-width: 600px; display: block; margin: auto;">
|
92 |
+
|
93 |
+
|
94 |
+
<div align="center">
|
95 |
+
<b style="font-size: 30px;">Unscripted video: live zero shot demonstration at story writing capabilities in Hebrew</b>
|
96 |
+
|
97 |
+
[![Zion_Alpha Story writing](https://img.youtube.com/vi/YYKeovnS0do/0.jpg)](https://www.youtube.com/watch?v=YYKeovnS0do)
|
98 |
+
</div>
|
99 |
+
|
100 |
+
<div align="center">
|
101 |
+
<b style="font-size: 30px;">Zion_Alpha VS Mistral 'Hebrew' Live & unscripted in real time</b>
|
102 |
+
|
103 |
+
[![Zion_Alpha Story writing](https://img.youtube.com/vi/YYKeovnS0do/0.jpg)](https://www.youtube.com/watch?v=DQFtx8M2txc)
|
104 |
+
</div>
|
105 |
+
|
106 |
+
<div align="center">
|
107 |
+
<b style="font-size: 30px;">Zion_Alpha VS Mistral 'Hebrew' Live & unscripted in real time Long text translation</b>
|
108 |
+
|
109 |
+
[![Zion_Alpha Story writing](https://img.youtube.com/vi/YYKeovnS0do/0.jpg)](https://www.youtube.com/watch?v=w5fz3Ot6tH8)
|
110 |
+
</div>
|
111 |
+
|
112 |
+
### History
|
113 |
+
The model was originally trained about 2 month after Mistral (v0.1) was released.
|
114 |
+
As of 04 June 2024, Zion_Alpha got the **Highest SNLI score in the world** among open source models in Hebrew, surpassing most of the models by a huge margin. (**84.05** score)
|
115 |
+
<img src="https://i.imgur.com/7HokS5w.png" alt="Zion_Alpha SNLI Score" style="width: 80%; min-width: 700px; display: block; margin: auto;">
|
116 |
+
|
117 |
+
### Support
|
118 |
+
<img src="https://i.imgur.com/0lHHN95.png" alt="GPUs too expensive" style="width: 10%; min-width: 100px; display: block; margin: left;">
|
119 |
+
|
120 |
+
- [My Ko-fi page](https://ko-fi.com/sicarius) ALL donations will go for research resources and compute, every bit counts ππ»
|
121 |
+
- [My Patreon](https://patreon.com/TenebraAI) ALL donations will go for research resources and compute, every bit counts ππ»
|
122 |
+
|