RichardErkhov
commited on
Commit
β’
ee73f65
1
Parent(s):
074f89d
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,119 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
CodeActAgent-Llama-2-7b - GGUF
|
11 |
+
- Model creator: https://huggingface.co/xingyaoww/
|
12 |
+
- Original model: https://huggingface.co/xingyaoww/CodeActAgent-Llama-2-7b/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [CodeActAgent-Llama-2-7b.Q2_K.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q2_K.gguf) | Q2_K | 2.36GB |
|
18 |
+
| [CodeActAgent-Llama-2-7b.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.IQ3_XS.gguf) | IQ3_XS | 2.6GB |
|
19 |
+
| [CodeActAgent-Llama-2-7b.IQ3_S.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.IQ3_S.gguf) | IQ3_S | 2.75GB |
|
20 |
+
| [CodeActAgent-Llama-2-7b.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q3_K_S.gguf) | Q3_K_S | 2.75GB |
|
21 |
+
| [CodeActAgent-Llama-2-7b.IQ3_M.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.IQ3_M.gguf) | IQ3_M | 2.9GB |
|
22 |
+
| [CodeActAgent-Llama-2-7b.Q3_K.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q3_K.gguf) | Q3_K | 3.07GB |
|
23 |
+
| [CodeActAgent-Llama-2-7b.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q3_K_M.gguf) | Q3_K_M | 3.07GB |
|
24 |
+
| [CodeActAgent-Llama-2-7b.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q3_K_L.gguf) | Q3_K_L | 3.35GB |
|
25 |
+
| [CodeActAgent-Llama-2-7b.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.IQ4_XS.gguf) | IQ4_XS | 3.4GB |
|
26 |
+
| [CodeActAgent-Llama-2-7b.Q4_0.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q4_0.gguf) | Q4_0 | 3.56GB |
|
27 |
+
| [CodeActAgent-Llama-2-7b.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.IQ4_NL.gguf) | IQ4_NL | 3.58GB |
|
28 |
+
| [CodeActAgent-Llama-2-7b.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q4_K_S.gguf) | Q4_K_S | 3.59GB |
|
29 |
+
| [CodeActAgent-Llama-2-7b.Q4_K.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q4_K.gguf) | Q4_K | 3.8GB |
|
30 |
+
| [CodeActAgent-Llama-2-7b.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q4_K_M.gguf) | Q4_K_M | 3.8GB |
|
31 |
+
| [CodeActAgent-Llama-2-7b.Q4_1.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q4_1.gguf) | Q4_1 | 3.95GB |
|
32 |
+
| [CodeActAgent-Llama-2-7b.Q5_0.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q5_0.gguf) | Q5_0 | 4.33GB |
|
33 |
+
| [CodeActAgent-Llama-2-7b.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q5_K_S.gguf) | Q5_K_S | 4.33GB |
|
34 |
+
| [CodeActAgent-Llama-2-7b.Q5_K.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q5_K.gguf) | Q5_K | 4.45GB |
|
35 |
+
| [CodeActAgent-Llama-2-7b.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q5_K_M.gguf) | Q5_K_M | 4.45GB |
|
36 |
+
| [CodeActAgent-Llama-2-7b.Q5_1.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q5_1.gguf) | Q5_1 | 4.72GB |
|
37 |
+
| [CodeActAgent-Llama-2-7b.Q6_K.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q6_K.gguf) | Q6_K | 5.15GB |
|
38 |
+
| [CodeActAgent-Llama-2-7b.Q8_0.gguf](https://huggingface.co/RichardErkhov/xingyaoww_-_CodeActAgent-Llama-2-7b-gguf/blob/main/CodeActAgent-Llama-2-7b.Q8_0.gguf) | Q8_0 | 6.67GB |
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
Original model description:
|
44 |
+
---
|
45 |
+
license: llama2
|
46 |
+
datasets:
|
47 |
+
- xingyaoww/code-act
|
48 |
+
language:
|
49 |
+
- en
|
50 |
+
tags:
|
51 |
+
- llm-agent
|
52 |
+
pipeline_tag: text-generation
|
53 |
+
---
|
54 |
+
|
55 |
+
<h1 align="center"> Executable Code Actions Elicit Better LLM Agents </h1>
|
56 |
+
|
57 |
+
<p align="center">
|
58 |
+
<a href="https://github.com/xingyaoww/code-act">π» Code</a>
|
59 |
+
β’
|
60 |
+
<a href="https://arxiv.org/abs/2402.01030">π Paper</a>
|
61 |
+
β’
|
62 |
+
<a href="https://huggingface.co/datasets/xingyaoww/code-act" >π€ Data (CodeActInstruct)</a>
|
63 |
+
β’
|
64 |
+
<a href="https://huggingface.co/xingyaoww/CodeActAgent-Mistral-7b-v0.1" >π€ Model (CodeActAgent-Mistral-7b-v0.1)</a>
|
65 |
+
β’
|
66 |
+
<a href="https://chat.xwang.dev/">π€ Chat with CodeActAgent!</a>
|
67 |
+
</p>
|
68 |
+
|
69 |
+
We propose to use executable Python **code** to consolidate LLM agentsβ **act**ions into a unified action space (**CodeAct**).
|
70 |
+
Integrated with a Python interpreter, CodeAct can execute code actions and dynamically revise prior actions or emit new actions upon new observations (e.g., code execution results) through multi-turn interactions.
|
71 |
+
|
72 |
+
![Overview](https://github.com/xingyaoww/code-act/blob/main/figures/overview.png?raw=true)
|
73 |
+
|
74 |
+
## Why CodeAct?
|
75 |
+
|
76 |
+
Our extensive analysis of 17 LLMs on API-Bank and a newly curated benchmark [M<sup>3</sup>ToolEval](docs/EVALUATION.md) shows that CodeAct outperforms widely used alternatives like Text and JSON (up to 20% higher success rate). Please check our paper for more detailed analysis!
|
77 |
+
|
78 |
+
![Comparison between CodeAct and Text/JSON](https://github.com/xingyaoww/code-act/blob/main/figures/codeact-comparison-table.png?raw=true)
|
79 |
+
*Comparison between CodeAct and Text / JSON as action.*
|
80 |
+
|
81 |
+
![Comparison between CodeAct and Text/JSON](https://github.com/xingyaoww/code-act/blob/main/figures/codeact-comparison-perf.png?raw=true)
|
82 |
+
*Quantitative results comparing CodeAct and {Text, JSON} on M<sup>3</sup>ToolEval.*
|
83 |
+
|
84 |
+
|
85 |
+
|
86 |
+
## π CodeActInstruct
|
87 |
+
|
88 |
+
We collect an instruction-tuning dataset CodeActInstruct that consists of 7k multi-turn interactions using CodeAct. Dataset is release at [huggingface dataset π€](https://huggingface.co/datasets/xingyaoww/code-act). Please refer to the paper and [this section](#-data-generation-optional) for details of data collection.
|
89 |
+
|
90 |
+
|
91 |
+
![Data Statistics](https://github.com/xingyaoww/code-act/blob/main/figures/data-stats.png?raw=true)
|
92 |
+
*Dataset Statistics. Token statistics are computed using Llama-2 tokenizer.*
|
93 |
+
|
94 |
+
## πͺ CodeActAgent
|
95 |
+
|
96 |
+
Trained on **CodeActInstruct** and general conversaions, **CodeActAgent** excels at out-of-domain agent tasks compared to open-source models of the same size, while not sacrificing generic performance (e.g., knowledge, dialog). We release two variants of CodeActAgent:
|
97 |
+
- **CodeActAgent-Mistral-7b-v0.1** (recommended, [model link](https://huggingface.co/xingyaoww/CodeActAgent-Mistral-7b-v0.1)): using Mistral-7b-v0.1 as the base model with 32k context window.
|
98 |
+
- **CodeActAgent-Llama-7b** ([model link](https://huggingface.co/xingyaoww/CodeActAgent-Llama-2-7b)): using Llama-2-7b as the base model with 4k context window.
|
99 |
+
|
100 |
+
![Model Performance](https://github.com/xingyaoww/code-act/blob/main/figures/model-performance.png?raw=true)
|
101 |
+
*Evaluation results for CodeActAgent. ID and OD stand for in-domain and out-of-domain evaluation correspondingly. Overall averaged performance normalizes the MT-Bench score to be consistent with other tasks and excludes in-domain tasks for fair comparison.*
|
102 |
+
|
103 |
+
|
104 |
+
Please check out [our paper](TODO) and [code](https://github.com/xingyaoww/code-act) for more details about data collection, model training, and evaluation.
|
105 |
+
|
106 |
+
|
107 |
+
## π Citation
|
108 |
+
|
109 |
+
```bibtex
|
110 |
+
@misc{wang2024executable,
|
111 |
+
title={Executable Code Actions Elicit Better LLM Agents},
|
112 |
+
author={Xingyao Wang and Yangyi Chen and Lifan Yuan and Yizhe Zhang and Yunzhu Li and Hao Peng and Heng Ji},
|
113 |
+
year={2024},
|
114 |
+
eprint={2402.01030},
|
115 |
+
archivePrefix={arXiv},
|
116 |
+
primaryClass={cs.CL}
|
117 |
+
}
|
118 |
+
```
|
119 |
+
|