maddes8cht
commited on
Commit
•
9c9f851
1
Parent(s):
de23514
"Update README.md"
Browse files
README.md
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
---
|
4 |
+
[![banner](https://maddes8cht.github.io/assets/buttons/Huggingface-banner.jpg)]()
|
5 |
+
|
6 |
+
## I am still building the structure of these descriptions.
|
7 |
+
|
8 |
+
These will contain increasingly more content to help find the best models for a purpose.
|
9 |
+
|
10 |
+
# openbuddy-falcon-7b-v6-bf16 - GGUF
|
11 |
+
- Model creator: [OpenBuddy](https://huggingface.co/OpenBuddy)
|
12 |
+
- Original model: [openbuddy-falcon-7b-v6-bf16](https://huggingface.co/OpenBuddy/openbuddy-falcon-7b-v6-bf16)
|
13 |
+
## Note:
|
14 |
+
|
15 |
+
This is v6 of OpenBuddy's Falcon-7b Variant. Somehow they forgot to provide a real `Model Card` for v6, so refer to the v5 `Model Card` instead:
|
16 |
+
|
17 |
+
https://huggingface.co/OpenBuddy/openbuddy-falcon-7b-v5-fp16
|
18 |
+
|
19 |
+
OpenBuddy provides strong multiligual Model variants. On their Huggingface Organization Card they say:
|
20 |
+
|
21 |
+
> Our mission with OpenBuddy is to provide a free, open, and offline-capable AI model that operates on users' devices, irrespective of their language or cultural background. We strive to empower individuals worldwide to access and benefit from AI technology.
|
22 |
+
|
23 |
+
# About GGUF format
|
24 |
+
|
25 |
+
`gguf` is the current file format used by the [`ggml`](https://github.com/ggerganov/ggml) library.
|
26 |
+
A growing list of Software is using it and can therefore use this model.
|
27 |
+
The core project making use of the ggml library is the [llama.cpp](https://github.com/ggerganov/llama.cpp) project by Georgi Gerganov
|
28 |
+
|
29 |
+
# Quantization variants
|
30 |
+
|
31 |
+
There is a bunch of quantized files available. How to choose the best for you:
|
32 |
+
|
33 |
+
# legacy quants
|
34 |
+
|
35 |
+
Q4_0, Q4_1, Q5_0, Q5_1 and Q8 are `legacy` quantization types.
|
36 |
+
Nevertheless, they are fully supported, as there are several circumstances that cause certain model not to be compatible with the modern K-quants.
|
37 |
+
Falcon 7B models cannot be quantized to K-quants.
|
38 |
+
|
39 |
+
# K-quants
|
40 |
+
|
41 |
+
K-quants are based on the idea that the quantization of certain parts affects the quality in different ways. If you quantize certain parts more and others less, you get a more powerful model with the same file size, or a smaller file size and lower memory load with comparable performance.
|
42 |
+
So, if possible, use K-quants.
|
43 |
+
With a Q6_K you should find it really hard to find a quality difference to the original model - ask your model two times the same question and you may encounter bigger quality differences.
|
44 |
+
|
45 |
+
|
46 |
+
# Original Model Card:
|
47 |
+
|
48 |
+
|
49 |
+
<center>
|
50 |
+
<a href="https://maddes8cht.github.com"><img src="/assets/buttons/maddes8cht-github-io.jpg" alt="GitHub" /></a>
|
51 |
+
<a href="https://stackexchange.com/users/26485911"><img src="https://stackexchange.com/users/flair/26485911.png" width="208" height="58" alt="profile for maddes8cht on Stack Exchange, a network of free, community-driven Q&A sites" title="profile for maddes8cht on Stack Exchange, a network of free, community-driven Q&A sites"></a>
|
52 |
+
<a href="https://github.com/maddes8cht"><img src="/assets/buttons/github-button.jpg" alt="GitHub" /></a>
|
53 |
+
<a href="https://huggingface.co/maddes8cht"><img src="/assets/buttons/huggingface-button.jpg" alt="HuggingFace" /></a></p>
|
54 |
+
<a href="https://twitter.com/maddes1966"><img src="/assets/buttons/twitter-button.jpg" alt="HuggingFace" /></a></p>
|
55 |
+
</center>
|