Commit
β’
57b3a6f
1
Parent(s):
7c2b9e7
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,136 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Quantization made by Richard Erkhov.
|
2 |
+
|
3 |
+
[Github](https://github.com/RichardErkhov)
|
4 |
+
|
5 |
+
[Discord](https://discord.gg/pvy7H8DZMG)
|
6 |
+
|
7 |
+
[Request more models](https://github.com/RichardErkhov/quant_request)
|
8 |
+
|
9 |
+
|
10 |
+
WizardLM-2-7B - bnb 4bits
|
11 |
+
- Model creator: https://huggingface.co/dreamgen/
|
12 |
+
- Original model: https://huggingface.co/dreamgen/WizardLM-2-7B/
|
13 |
+
|
14 |
+
|
15 |
+
|
16 |
+
|
17 |
+
Original model description:
|
18 |
+
---
|
19 |
+
license: apache-2.0
|
20 |
+
---
|
21 |
+
|
22 |
+
|
23 |
+
|
24 |
+
|
25 |
+
<p style="font-size:20px;" align="center">
|
26 |
+
π <a href="https://wizardlm.github.io/WizardLM2" target="_blank">WizardLM-2 Release Blog</a> </p>
|
27 |
+
<p align="center">
|
28 |
+
π€ <a href="https://huggingface.co/collections/microsoft/wizardlm-2-661d403f71e6c8257dbd598a" target="_blank">HF Repo</a> β’π± <a href="https://github.com/victorsungo/WizardLM/tree/main/WizardLM-2" target="_blank">Github Repo</a> β’ π¦ <a href="https://twitter.com/WizardLM_AI" target="_blank">Twitter</a> β’ π <a href="https://arxiv.org/abs/2304.12244" target="_blank">[WizardLM]</a> β’ π <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> β’ π <a href="https://arxiv.org/abs/2308.09583" target="_blank">[WizardMath]</a> <br>
|
29 |
+
</p>
|
30 |
+
<p align="center">
|
31 |
+
π Join our <a href="https://discord.gg/VZjjHtWrKs" target="_blank">Discord</a>
|
32 |
+
</p>
|
33 |
+
|
34 |
+
|
35 |
+
|
36 |
+
## News π₯π₯π₯ [2024/04/15]
|
37 |
+
|
38 |
+
We introduce and opensource WizardLM-2, our next generation state-of-the-art large language models,
|
39 |
+
which have improved performance on complex chat, multilingual, reasoning and agent.
|
40 |
+
New family includes three cutting-edge models: WizardLM-2 8x22B, WizardLM-2 70B, and WizardLM-2 7B.
|
41 |
+
|
42 |
+
- WizardLM-2 8x22B is our most advanced model, demonstrates highly competitive performance compared to those leading proprietary works
|
43 |
+
and consistently outperforms all the existing state-of-the-art opensource models.
|
44 |
+
- WizardLM-2 70B reaches top-tier reasoning capabilities and is the first choice in the same size.
|
45 |
+
- WizardLM-2 7B is the fastest and achieves comparable performance with existing 10x larger opensource leading models.
|
46 |
+
|
47 |
+
For more details of WizardLM-2 please read our [release blog post](https://wizardlm.github.io/WizardLM2) and upcoming paper.
|
48 |
+
|
49 |
+
|
50 |
+
## Model Details
|
51 |
+
|
52 |
+
* **Model name**: WizardLM-2 7B
|
53 |
+
* **Developed by**: WizardLM@Microsoft AI
|
54 |
+
* **Base model**: [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1)
|
55 |
+
* **Parameters**: 7B
|
56 |
+
* **Language(s)**: Multilingual
|
57 |
+
* **Blog**: [Introducing WizardLM-2](https://wizardlm.github.io/WizardLM2)
|
58 |
+
* **Repository**: [https://github.com/nlpxucan/WizardLM](https://github.com/nlpxucan/WizardLM)
|
59 |
+
* **Paper**: WizardLM-2 (Upcoming)
|
60 |
+
* **License**: Apache2.0
|
61 |
+
|
62 |
+
|
63 |
+
|
64 |
+
## Model Capacities
|
65 |
+
|
66 |
+
|
67 |
+
**MT-Bench**
|
68 |
+
|
69 |
+
We also adopt the automatic MT-Bench evaluation framework based on GPT-4 proposed by lmsys to assess the performance of models.
|
70 |
+
The WizardLM-2 8x22B even demonstrates highly competitive performance compared to the most advanced proprietary models.
|
71 |
+
Meanwhile, WizardLM-2 7B and WizardLM-2 70B are all the top-performing models among the other leading baselines at 7B to 70B model scales.
|
72 |
+
|
73 |
+
<p align="center" width="100%">
|
74 |
+
<a ><img src="https://raw.githubusercontent.com/WizardLM/WizardLM2/main/static/images/mtbench.png" alt="MTBench" style="width: 96%; min-width: 300px; display: block; margin: auto;"></a>
|
75 |
+
</p>
|
76 |
+
|
77 |
+
|
78 |
+
**Human Preferences Evaluation**
|
79 |
+
|
80 |
+
We carefully collected a complex and challenging set consisting of real-world instructions, which includes main requirements of humanity, such as writing, coding, math, reasoning, agent, and multilingual.
|
81 |
+
We report the win:loss rate without tie:
|
82 |
+
|
83 |
+
- WizardLM-2 8x22B is just slightly falling behind GPT-4-1106-preview, and significantly stronger than Command R Plus and GPT4-0314.
|
84 |
+
- WizardLM-2 70B is better than GPT4-0613, Mistral-Large, and Qwen1.5-72B-Chat.
|
85 |
+
- WizardLM-2 7B is comparable with Qwen1.5-32B-Chat, and surpasses Qwen1.5-14B-Chat and Starling-LM-7B-beta.
|
86 |
+
|
87 |
+
<p align="center" width="100%">
|
88 |
+
<a ><img src="https://raw.githubusercontent.com/WizardLM/WizardLM2/main/static/images/winall.png" alt="Win" style="width: 96%; min-width: 300px; display: block; margin: auto;"></a>
|
89 |
+
</p>
|
90 |
+
|
91 |
+
|
92 |
+
|
93 |
+
|
94 |
+
|
95 |
+
## Method Overview
|
96 |
+
We built a **fully AI powered synthetic training system** to train WizardLM-2 models, please refer to our [blog](https://wizardlm.github.io/WizardLM2) for more details of this system.
|
97 |
+
|
98 |
+
<p align="center" width="100%">
|
99 |
+
<a ><img src="https://raw.githubusercontent.com/WizardLM/WizardLM2/main/static/images/exp_1.png" alt="Method" style="width: 96%; min-width: 300px; display: block; margin: auto;"></a>
|
100 |
+
</p>
|
101 |
+
|
102 |
+
|
103 |
+
|
104 |
+
|
105 |
+
|
106 |
+
## Usage
|
107 |
+
|
108 |
+
β<b>Note for model system prompts usage:</b>
|
109 |
+
|
110 |
+
|
111 |
+
<b>WizardLM-2</b> adopts the prompt format from <b>Vicuna</b> and supports **multi-turn** conversation. The prompt should be as following:
|
112 |
+
|
113 |
+
```
|
114 |
+
A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful,
|
115 |
+
detailed, and polite answers to the user's questions. USER: Hi ASSISTANT: Hello.</s>
|
116 |
+
USER: Who are you? ASSISTANT: I am WizardLM.</s>......
|
117 |
+
```
|
118 |
+
|
119 |
+
<b> Inference WizardLM-2 Demo Script</b>
|
120 |
+
|
121 |
+
We provide a WizardLM-2 inference demo [code](https://github.com/nlpxucan/WizardLM/tree/main/demo) on our github.
|
122 |
+
|
123 |
+
|
124 |
+
|
125 |
+
|
126 |
+
|
127 |
+
|
128 |
+
|
129 |
+
|
130 |
+
|
131 |
+
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
|
136 |
+
|