rAIfle commited on
Commit
f84cce0
β€’
1 Parent(s): 6b21091

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +141 -0
README.md ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ ---
4
+
5
+ ```
6
+ e88 88e d8
7
+ d888 888b 8888 8888 ,"Y88b 888 8e d88
8
+ C8888 8888D 8888 8888 "8" 888 888 88b d88888
9
+ Y888 888P Y888 888P ,ee 888 888 888 888
10
+ "88 88" "88 88" "88 888 888 888 888
11
+ b
12
+ 8b,
13
+
14
+ e88'Y88 d8 888
15
+ d888 'Y ,"Y88b 888,8, d88 ,e e, 888
16
+ C8888 "8" 888 888 " d88888 d88 88b 888
17
+ Y888 ,d ,ee 888 888 888 888 , 888
18
+ "88,d88 "88 888 888 888 "YeeP" 888
19
+
20
+ PROUDLY PRESENTS
21
+ ```
22
+ # WizardLM-2-8x22B-exl2-rpcal
23
+
24
+ Quantized using 200 samples of 8192 tokens from an RP-oriented [PIPPA](https://huggingface.co/datasets/royallab/PIPPA-cleaned) dataset.
25
+
26
+ Branches:
27
+ - `main` -- `measurement.json`
28
+ - `4.5b6h` -- 4.5bpw, 6bit lm_head
29
+ - `4b6h` -- 4bpw, 6bit lm_head
30
+
31
+ Original model link: (reuploaded, original source got taken down) [alpindale/WizardLM-2-8x22B](https://huggingface.co/alpindale/WizardLM-2-8x22B)
32
+
33
+ ### Quanter's notes
34
+ I like this. On the `main`-branch, I added the settings I use in ST, but there are a few "moving parts" here. I switch sysprompt between the massive one I include, Crackhead Agent 47, and the old classic included with simple-proxy.
35
+
36
+ Original model README below.
37
+
38
+ -----
39
+
40
+
41
+ <p style="font-size:20px;" align="center">
42
+ 🏠 <a href="https://wizardlm.github.io/WizardLM2" target="_blank">WizardLM-2 Release Blog</a> </p>
43
+ <p align="center">
44
+ πŸ€— <a href="https://huggingface.co/collections/microsoft/wizardlm-2-661d403f71e6c8257dbd598a" target="_blank">HF Repo</a> β€’πŸ± <a href="https://github.com/victorsungo/WizardLM/tree/main/WizardLM-2" target="_blank">Github Repo</a> β€’ 🐦 <a href="https://twitter.com/WizardLM_AI" target="_blank">Twitter</a> β€’ πŸ“ƒ <a href="https://arxiv.org/abs/2304.12244" target="_blank">[WizardLM]</a> β€’ πŸ“ƒ <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> β€’ πŸ“ƒ <a href="https://arxiv.org/abs/2308.09583" target="_blank">[WizardMath]</a> <br>
45
+ </p>
46
+ <p align="center">
47
+ πŸ‘‹ Join our <a href="https://discord.gg/VZjjHtWrKs" target="_blank">Discord</a>
48
+ </p>
49
+
50
+ ## See [here](https://huggingface.co/lucyknada/microsoft_WizardLM-2-7B) for the WizardLM-2-7B re-upload.
51
+
52
+ ## News πŸ”₯πŸ”₯πŸ”₯ [2024/04/15]
53
+
54
+ We introduce and opensource WizardLM-2, our next generation state-of-the-art large language models,
55
+ which have improved performance on complex chat, multilingual, reasoning and agent.
56
+ New family includes three cutting-edge models: WizardLM-2 8x22B, WizardLM-2 70B, and WizardLM-2 7B.
57
+
58
+ - WizardLM-2 8x22B is our most advanced model, demonstrates highly competitive performance compared to those leading proprietary works
59
+ and consistently outperforms all the existing state-of-the-art opensource models.
60
+ - WizardLM-2 70B reaches top-tier reasoning capabilities and is the first choice in the same size.
61
+ - WizardLM-2 7B is the fastest and achieves comparable performance with existing 10x larger opensource leading models.
62
+
63
+ For more details of WizardLM-2 please read our [release blog post](https://web.archive.org/web/20240415221214/https://wizardlm.github.io/WizardLM2/) and upcoming paper.
64
+
65
+
66
+ ## Model Details
67
+
68
+ * **Model name**: WizardLM-2 8x22B
69
+ * **Developed by**: WizardLM@Microsoft AI
70
+ * **Model type**: Mixture of Experts (MoE)
71
+ * **Base model**: [mistral-community/Mixtral-8x22B-v0.1](https://huggingface.co/mistral-community/Mixtral-8x22B-v0.1)
72
+ * **Parameters**: 141B
73
+ * **Language(s)**: Multilingual
74
+ * **Blog**: [Introducing WizardLM-2](https://web.archive.org/web/20240415221214/https://wizardlm.github.io/WizardLM2/)
75
+ * **Repository**: [https://github.com/nlpxucan/WizardLM](https://github.com/nlpxucan/WizardLM)
76
+ * **Paper**: WizardLM-2 (Upcoming)
77
+ * **License**: Apache2.0
78
+
79
+
80
+ ## Model Capacities
81
+
82
+
83
+ **MT-Bench**
84
+
85
+ We also adopt the automatic MT-Bench evaluation framework based on GPT-4 proposed by lmsys to assess the performance of models.
86
+ The WizardLM-2 8x22B even demonstrates highly competitive performance compared to the most advanced proprietary models.
87
+ 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.
88
+
89
+ <p align="center" width="100%">
90
+ <a ><img src="https://web.archive.org/web/20240415175608im_/https://wizardlm.github.io/WizardLM2/static/images/mtbench.png" alt="MTBench" style="width: 96%; min-width: 300px; display: block; margin: auto;"></a>
91
+ </p>
92
+
93
+
94
+ **Human Preferences Evaluation**
95
+
96
+ 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.
97
+ We report the win:loss rate without tie:
98
+
99
+ - WizardLM-2 8x22B is just slightly falling behind GPT-4-1106-preview, and significantly stronger than Command R Plus and GPT4-0314.
100
+ - WizardLM-2 70B is better than GPT4-0613, Mistral-Large, and Qwen1.5-72B-Chat.
101
+ - WizardLM-2 7B is comparable with Qwen1.5-32B-Chat, and surpasses Qwen1.5-14B-Chat and Starling-LM-7B-beta.
102
+
103
+ <p align="center" width="100%">
104
+ <a ><img src="https://web.archive.org/web/20240415163303im_/https://wizardlm.github.io/WizardLM2/static/images/winall.png" alt="Win" style="width: 96%; min-width: 300px; display: block; margin: auto;"></a>
105
+ </p>
106
+
107
+
108
+
109
+
110
+
111
+ ## Method Overview
112
+ We built a **fully AI powered synthetic training system** to train WizardLM-2 models, please refer to our [blog](https://web.archive.org/web/20240415221214/https://wizardlm.github.io/WizardLM2/) for more details of this system.
113
+
114
+ <p align="center" width="100%">
115
+ <a ><img src="https://web.archive.org/web/20240415163303im_/https://wizardlm.github.io/WizardLM2/static/images/exp_1.png" alt="Method" style="width: 96%; min-width: 300px; display: block; margin: auto;"></a>
116
+ </p>
117
+
118
+
119
+
120
+
121
+
122
+ ## Usage
123
+
124
+ ❗<b>Note for model system prompts usage:</b>
125
+
126
+
127
+ <b>WizardLM-2</b> adopts the prompt format from <b>Vicuna</b> and supports **multi-turn** conversation. The prompt should be as following:
128
+
129
+ ```
130
+ A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful,
131
+ detailed, and polite answers to the user's questions. USER: Hi ASSISTANT: Hello.</s>
132
+ USER: Who are you? ASSISTANT: I am WizardLM.</s>......
133
+ ```
134
+
135
+ <b> Inference WizardLM-2 Demo Script</b>
136
+
137
+ We provide a WizardLM-2 inference demo [code](https://github.com/nlpxucan/WizardLM/tree/main/demo) on our github.
138
+
139
+
140
+
141
+