Update README.md
Browse files
README.md
CHANGED
@@ -2,4 +2,52 @@
|
|
2 |
license: apache-2.0
|
3 |
---
|
4 |
|
5 |
-
This is experimental model, yet it is the Most powerful RNN model in the world.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: apache-2.0
|
3 |
---
|
4 |
|
5 |
+
This is a experimental model, yet it is the Most powerful RNN model in the world.
|
6 |
+
|
7 |
+
# Mobius RWKV r6 chat 12B 16k
|
8 |
+
Mobius is a RWKV v6 arch chat model, benifit from [Matrix-Valued States and Dynamic Recurrence](https://arxiv.org/abs/2404.05892)
|
9 |
+
|
10 |
+
## Introduction
|
11 |
+
|
12 |
+
Mobius is a RWKV v6 arch model, a state based RNN+CNN+Transformer Mixed language model pretrained on a certain amount of data.
|
13 |
+
In comparison with the previous released Mobius, the improvements include:
|
14 |
+
|
15 |
+
* Only 24G Vram to run this model locally with fp16;
|
16 |
+
* Significant performance improvement;
|
17 |
+
* Multilingual support ;
|
18 |
+
* Stable support of 16K context length.
|
19 |
+
* function call support ;
|
20 |
+
|
21 |
+
|
22 |
+
## Usage
|
23 |
+
We encourage you use few shots to use this model, Desipte Directly use User: xxxx\n\nAssistant: xxx\n\n is really good too, Can boost all potential ability.
|
24 |
+
|
25 |
+
Recommend Temp and topp: 0.7 0.6/1 0.3/1.5 0.3/0.2 0.8
|
26 |
+
|
27 |
+
function call format:
|
28 |
+
System: func xxxx
|
29 |
+
|
30 |
+
User: xxxx
|
31 |
+
|
32 |
+
Assistant: xxxx
|
33 |
+
|
34 |
+
Obersavtion: xxxx
|
35 |
+
|
36 |
+
Assistant: xxxx
|
37 |
+
|
38 |
+
## More details
|
39 |
+
Mobius 12B 16k based on RWKV v6 arch, which is leading state based RNN+CNN+Transformer Mixed large language model which focus opensouce community
|
40 |
+
* 10~100 trainning/inference cost reduce;
|
41 |
+
* state based,selected memory, which mean good at grok;
|
42 |
+
* community support.
|
43 |
+
|
44 |
+
## requirements
|
45 |
+
24G vram to run fp16, 12G for int8, 6G for nf4 with Ai00 server.
|
46 |
+
|
47 |
+
* [RWKV Runner](https://github.com/josStorer/RWKV-Runner)
|
48 |
+
* [Ai00 server](https://github.com/cgisky1980/ai00_rwkv_server)
|
49 |
+
|
50 |
+
## Benchmark
|
51 |
+
Ceval:
|
52 |
+
Cmmlu:
|
53 |
+
|