azhx commited on
Commit
8bce553
1 Parent(s): 9a780c7

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +116 -0
README.md CHANGED
@@ -1,3 +1,119 @@
1
  ---
2
  license: mit
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: mit
3
+ datasets:
4
+ - TIGER-Lab/SKGInstruct
5
+ language:
6
+ - en
7
  ---
8
+ # 🏗️ StructLM: Towards Building Generalist Models for Structured Knowledge Grounding
9
+
10
+
11
+
12
+ Project Page: [https://tiger-ai-lab.github.io/StructLM/](https://tiger-ai-lab.github.io/StructLM/)
13
+
14
+ Paper: [https://arxiv.org/pdf/2402.16671.pdf](https://arxiv.org/pdf/2402.16671.pdf)
15
+
16
+ Code: [https://github.com/TIGER-AI-Lab/StructLM](https://github.com/TIGER-AI-Lab/StructLM)
17
+
18
+
19
+ ![Alt text](https://raw.githubusercontent.com/TIGER-AI-Lab/StructLM/gh-pages/static/images/thumbnail.drawio.png)
20
+
21
+
22
+ ## Introduction
23
+ StructLM, is a series of open-source large language models (LLMs) finetuned for structured knowledge grounding (SKG) tasks.
24
+
25
+ This model is trained using Mistral as the base model, instead of CodeLlama.
26
+
27
+ ## Training Data
28
+ This model is trained on 🤗 [SKGInstruct-skg-only Dataset](https://huggingface.co/datasets/TIGER-Lab/SKGInstruct-skg-only). Check out the dataset card for more details.
29
+
30
+ ## Training Procedure
31
+ The models are fine-tuned with CodeLlama-Instruct-hf models as base models. Each model is trained for 3 epochs, and the best checkpoint is selected.
32
+
33
+ ## Evaluation
34
+ Reference our project page for evaluation results on 7B-M
35
+
36
+ ## Usage
37
+ You can use the models through Huggingface's Transformers library.
38
+ Check our Github repo for the evaluation code: [https://github.com/TIGER-AI-Lab/StructLM](https://github.com/TIGER-AI-Lab/StructLM)
39
+
40
+
41
+ ## Prompt Format
42
+
43
+ **For this 7B model, the prompt format is**
44
+ ```
45
+ Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
46
+
47
+
48
+ ### Instruction:
49
+ {instruction}
50
+
51
+ {input}
52
+
53
+ {question}
54
+
55
+ ### Response:
56
+
57
+ ```
58
+
59
+ To see concrete examples of this linearization, you can directly reference the 🤗 [SKGInstruct Dataset](https://huggingface.co/datasets/TIGER-Lab/SKGInstruct) (coming soon).
60
+ We will provide code for linearizing this data shortly.
61
+
62
+
63
+ A few examples:
64
+
65
+ **Tabular data**
66
+ ```
67
+ col : day | kilometers row 1 : tuesday | 0 row 2 : wednesday | 0 row 3 : thursday | 4 row 4 : friday | 0 row 5 : saturday | 0
68
+ ```
69
+
70
+ **Knowledge triples (dart)**
71
+ ```
72
+ Hawaii Five-O : notes : Episode: The Flight of the Jewels | [TABLECONTEXT] : [title] : Jeff Daniels | [TABLECONTEXT] : title : Hawaii Five-O
73
+ ```
74
+
75
+ **Knowledge graph schema (grailqa)**
76
+ ```
77
+ top antiquark: m.094nrqp | physics.particle_antiparticle.self_antiparticle physics.particle_family physics.particle.antiparticle physics.particle_family.subclasses physics.subatomic_particle_generation physics.particle_family.particles physics.particle common.image.appears_in_topic_gallery physics.subatomic_particle_generation.particles physics.particle.family physics.particle_family.parent_class physics.particle_antiparticle physics.particle_antiparticle.particle physics.particle.generation
78
+ ```
79
+
80
+ **Example input**
81
+
82
+ ```
83
+ Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
84
+
85
+
86
+ ### Instruction:
87
+ Use the information in the following table to solve the problem, choose between the choices if they are provided. table:
88
+
89
+ col : day | kilometers row 1 : tuesday | 0 row 2 : wednesday | 0 row 3 : thursday | 4 row 4 : friday | 0 row 5 : saturday | 0
90
+
91
+ question:
92
+
93
+ Allie kept track of how many kilometers she walked during the past 5 days. What is the range of the numbers? [/INST]
94
+
95
+ ### Response:
96
+
97
+ ```
98
+
99
+
100
+ ## Intended Uses
101
+ These models are trained for research purposes. They are designed to be proficient in interpreting linearized structured input. Downstream uses can potentially include various applications requiring the interpretation of structured data.
102
+
103
+ ## Limitations
104
+ While we've tried to build an SKG-specialized model capable of generalizing, we have shown that this is a challenging domain, and it may lack performance characteristics that allow it to be directly used in chat or other applications.
105
+
106
+
107
+ ## Citation
108
+ If you use the models, data, or code from this project, please cite the original paper:
109
+
110
+ ```
111
+ @misc{zhuang2024structlm,
112
+ title={StructLM: Towards Building Generalist Models for Structured Knowledge Grounding},
113
+ author={Alex Zhuang and Ge Zhang and Tianyu Zheng and Xinrun Du and Junjie Wang and Weiming Ren and Stephen W. Huang and Jie Fu and Xiang Yue and Wenhu Chen},
114
+ year={2024},
115
+ eprint={2402.16671},
116
+ archivePrefix={arXiv},
117
+ primaryClass={cs.CL}
118
+ }
119
+ ```