File size: 1,978 Bytes
339f0ef
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92d74a9
339f0ef
 
 
 
 
d523cb8
339f0ef
f8dfc57
339f0ef
34a1c74
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
339f0ef
 
 
 
 
 
 
34a1c74
 
339f0ef
 
 
 
 
 
 
56fd7b3
339f0ef
 
 
 
 
3524a7e
 
339f0ef
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
---
license: apache-2.0
---

This is a experimental model, yet it is the Most powerful RNN model in the world.

# Mobius RWKV r6 chat 12B 16k
Mobius is a RWKV v6 arch chat model, benifit from [Matrix-Valued States and Dynamic Recurrence](https://arxiv.org/abs/2404.05892)

## Introduction

Mobius is a RWKV v6 arch model, a state based RNN+CNN+Transformer Mixed language model pretrained on a certain amount of data.
In comparison with the previous released Mobius, the improvements include:

* Only 24G Vram to run this model locally with fp16;
* Significant performance improvement in chinese;
* Stable support of 16K context length.
* function call support ;
  

## Usage
Chat format: User: xxxx\n\nAssistant: xxx\n\n

Recommend Temp and topp: 1 0.3

function call format example: 

```
System: You are a helpful assistant with access to the following functions. Use them if required -{
  "name": "get_exchange_rate",
  "description": "Get the exchange rate between two currencies",
  "parameters": {
    "type": "object",
    "properties": {
    "base_currency": {
    "type": "string",
    "description": "The currency to convert from"
    },
    "target_currency": {
    "type": "string",
    "description": "The currency to convert to"
    }
    },
    "required": [
    "base_currency",
    "target_currency"
    ]
  }
}

User: Hi, I need to know the exchange rate from USD to EUR

Assistant: xxxx

Obersavtion: xxxx

Assistant: xxxx

```

## More details
Mobius 12B 16k based on RWKV v6 arch, which is leading state based RNN+CNN+Transformer Mixed large language model which focus opensouce community
* 10~100 trainning/inference cost reduce;
* state based,selected memory, which mean good at grok;
* community support.

## requirements
21.9G vram to run fp16, 13.7G for int8, 7.2 for nf4 with Ai00 server.

* [RWKV Runner](https://github.com/josStorer/RWKV-Runner)
* [Ai00 server](https://github.com/cgisky1980/ai00_rwkv_server)

## Benchmark
ceval 63.53
cmmlu 76.07