Add missing metadata and improve formatting (#1)
Browse files- Add missing metadata and improve formatting (8cddae10628314a8b0121d3d3a7168d2cc74b022)
Co-authored-by: Niels Rogge <[email protected]>
README.md
CHANGED
|
@@ -1,3 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
Quantization made by Richard Erkhov.
|
| 2 |
|
| 3 |
[Github](https://github.com/RichardErkhov)
|
|
@@ -6,23 +12,12 @@ Quantization made by Richard Erkhov.
|
|
| 6 |
|
| 7 |
[Request more models](https://github.com/RichardErkhov/quant_request)
|
| 8 |
|
| 9 |
-
|
| 10 |
CursorCore-Yi-1.5B-SR - AWQ
|
| 11 |
- Model creator: https://huggingface.co/TechxGenus/
|
| 12 |
- Original model: https://huggingface.co/TechxGenus/CursorCore-Yi-1.5B-SR/
|
| 13 |
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
Original model description:
|
| 18 |
-
|
| 19 |
-
tags:
|
| 20 |
-
- code
|
| 21 |
-
base_model:
|
| 22 |
-
- 01-ai/Yi-Coder-1.5B
|
| 23 |
-
library_name: transformers
|
| 24 |
-
pipeline_tag: text-generation
|
| 25 |
-
license: apache-2.0
|
| 26 |
---
|
| 27 |
|
| 28 |
# CursorCore: Assist Programming through Aligning Anything
|
|
@@ -32,14 +27,15 @@ license: apache-2.0
|
|
| 32 |
<a href="https://hf.co/papers/2410.07002">[🤗HF Paper]</a> |
|
| 33 |
<a href="https://huggingface.co/collections/TechxGenus/cursorcore-series-6706618c38598468866b60e2">[🤖Models]</a> |
|
| 34 |
<a href="https://github.com/TechxGenus/CursorCore">[🛠️Code]</a> |
|
| 35 |
-
<a href="https://github.com/TechxGenus/CursorWeb">[Web]</a> |
|
| 36 |
-
<a href="https://discord.gg/Z5Tev8fV">[Discord]</a>
|
| 37 |
</p>
|
| 38 |
|
| 39 |
<hr>
|
| 40 |
|
| 41 |
- [CursorCore: Assist Programming through Aligning Anything](#cursorcore-assist-programming-through-aligning-anything)
|
| 42 |
- [Introduction](#introduction)
|
|
|
|
| 43 |
- [Models](#models)
|
| 44 |
- [Usage](#usage)
|
| 45 |
- [1) Normal chat](#1-normal-chat)
|
|
@@ -47,6 +43,7 @@ license: apache-2.0
|
|
| 47 |
- [3) Web Demo](#3-web-demo)
|
| 48 |
- [Future Work](#future-work)
|
| 49 |
- [Citation](#citation)
|
|
|
|
| 50 |
- [Contribution](#contribution)
|
| 51 |
|
| 52 |
<hr>
|
|
@@ -56,15 +53,43 @@ license: apache-2.0
|
|
| 56 |
CursorCore is a series of open-source models designed for AI-assisted programming. It aims to support features such as automated editing and inline chat, replicating the core abilities of closed-source AI-assisted programming tools like Cursor. This is achieved by aligning data generated through Programming-Instruct. Please read [our paper](http://arxiv.org/abs/2410.07002) to learn more.
|
| 57 |
|
| 58 |
<p align="center">
|
| 59 |
-
<img width="100%" alt="conversation" src="https://
|
| 60 |
</p>
|
| 61 |
|
| 62 |
-
. We also provide pre-quantized weights for GPTQ and AWQ here: [CursorCore-Quantization](https://huggingface.co/collections/TechxGenus/cursorcore-quantization-67066431f29f252494ee8cf3)
|
| 67 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
## Usage
|
| 69 |
|
| 70 |
Here are some examples of how to use our model:
|
|
@@ -131,13 +156,27 @@ sample = {
|
|
| 131 |
{
|
| 132 |
"type": "code",
|
| 133 |
"lang": "python",
|
| 134 |
-
"code": """def quick_sort(arr)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 135 |
}
|
| 136 |
],
|
| 137 |
"current": {
|
| 138 |
"type": "code",
|
| 139 |
"lang": "python",
|
| 140 |
-
"code": """def quick_sort(array)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 141 |
},
|
| 142 |
"user": ""
|
| 143 |
}
|
|
@@ -219,7 +258,14 @@ sample = {
|
|
| 219 |
"current": {
|
| 220 |
"type": "code",
|
| 221 |
"lang": "python",
|
| 222 |
-
"code": """def quick_sort(array)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 223 |
},
|
| 224 |
"user": "Add Docstring."
|
| 225 |
}
|
|
@@ -290,7 +336,14 @@ sample = {
|
|
| 290 |
"current": {
|
| 291 |
"type": "code",
|
| 292 |
"lang": "python",
|
| 293 |
-
"code": """def quick_sort(array)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 294 |
},
|
| 295 |
"user": "Add Docstring."
|
| 296 |
}
|
|
@@ -359,7 +412,14 @@ sample = {
|
|
| 359 |
"current": {
|
| 360 |
"type": "code",
|
| 361 |
"lang": "python",
|
| 362 |
-
"code": """def quick_sort(array)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 363 |
},
|
| 364 |
"user": "Add Docstring."
|
| 365 |
}
|
|
@@ -429,8 +489,12 @@ CursorCore is still in a very early stage, and lots of work is needed to achieve
|
|
| 429 |
}
|
| 430 |
```
|
| 431 |
|
| 432 |
-
##
|
|
|
|
|
|
|
| 433 |
|
| 434 |
-
|
| 435 |
|
|
|
|
| 436 |
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
library_name: transformers
|
| 4 |
+
pipeline_tag: text-generation
|
| 5 |
+
---
|
| 6 |
+
|
| 7 |
Quantization made by Richard Erkhov.
|
| 8 |
|
| 9 |
[Github](https://github.com/RichardErkhov)
|
|
|
|
| 12 |
|
| 13 |
[Request more models](https://github.com/RichardErkhov/quant_request)
|
| 14 |
|
|
|
|
| 15 |
CursorCore-Yi-1.5B-SR - AWQ
|
| 16 |
- Model creator: https://huggingface.co/TechxGenus/
|
| 17 |
- Original model: https://huggingface.co/TechxGenus/CursorCore-Yi-1.5B-SR/
|
| 18 |
|
|
|
|
|
|
|
|
|
|
| 19 |
Original model description:
|
| 20 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
---
|
| 22 |
|
| 23 |
# CursorCore: Assist Programming through Aligning Anything
|
|
|
|
| 27 |
<a href="https://hf.co/papers/2410.07002">[🤗HF Paper]</a> |
|
| 28 |
<a href="https://huggingface.co/collections/TechxGenus/cursorcore-series-6706618c38598468866b60e2">[🤖Models]</a> |
|
| 29 |
<a href="https://github.com/TechxGenus/CursorCore">[🛠️Code]</a> |
|
| 30 |
+
<a href="https://github.com/TechxGenus/CursorWeb">[<img src="https://github.com/TechxGenus/CursorCore/blob/main/pictures/cursorcore.png" width="12.5px">Web]</a> |
|
| 31 |
+
<a href="https://discord.gg/Z5Tev8fV">[<img src="https://github.com/TechxGenus/CursorCore/blob/main/pictures/discord.png" width="15x">Discord]</a>
|
| 32 |
</p>
|
| 33 |
|
| 34 |
<hr>
|
| 35 |
|
| 36 |
- [CursorCore: Assist Programming through Aligning Anything](#cursorcore-assist-programming-through-aligning-anything)
|
| 37 |
- [Introduction](#introduction)
|
| 38 |
+
- [Structure](#structure)
|
| 39 |
- [Models](#models)
|
| 40 |
- [Usage](#usage)
|
| 41 |
- [1) Normal chat](#1-normal-chat)
|
|
|
|
| 43 |
- [3) Web Demo](#3-web-demo)
|
| 44 |
- [Future Work](#future-work)
|
| 45 |
- [Citation](#citation)
|
| 46 |
+
- [Acknowledgements](#acknowledgements)
|
| 47 |
- [Contribution](#contribution)
|
| 48 |
|
| 49 |
<hr>
|
|
|
|
| 53 |
CursorCore is a series of open-source models designed for AI-assisted programming. It aims to support features such as automated editing and inline chat, replicating the core abilities of closed-source AI-assisted programming tools like Cursor. This is achieved by aligning data generated through Programming-Instruct. Please read [our paper](http://arxiv.org/abs/2410.07002) to learn more.
|
| 54 |
|
| 55 |
<p align="center">
|
| 56 |
+
<img width="100%" alt="conversation" src="https://github.com/TechxGenus/CursorCore/blob/main/pictures/conversation.png">
|
| 57 |
</p>
|
| 58 |
|
| 59 |
+

|
| 60 |
+
|
| 61 |
+
## Structure
|
| 62 |
+
|
| 63 |
+
- `./benchmark` contains the APEval benchmark
|
| 64 |
+
- `./data` contains code to preprocess datasets
|
| 65 |
+
- `./eval` contains code to evaluate models
|
| 66 |
+
- `./gen` contains code to prompt LLMs for generation
|
| 67 |
+
- `./generic` common functions, tools and special tokens
|
| 68 |
+
- `./src` contains code about Programming-Instruct
|
| 69 |
+
- `./train` contains code for training CursorCore
|
| 70 |
+
|
| 71 |
+
Please ensure all dependencies are installed using the following command:
|
| 72 |
+
|
| 73 |
+
```bash
|
| 74 |
+
pip install -r requirements.txt
|
| 75 |
+
```
|
| 76 |
+
|
| 77 |
+
We also use [flash-attention](https://github.com/Dao-AILab/flash-attention) for efficient training and [flashinfer](https://github.com/flashinfer-ai/flashinfer) to accelerate inference. See the documents for them to learn how to install.
|
| 78 |
|
| 79 |
## Models
|
| 80 |
|
| 81 |
Our models have been open-sourced on Hugging Face. You can access our models here: [CursorCore-Series](https://huggingface.co/collections/TechxGenus/cursorcore-series-6706618c38598468866b60e2"). We also provide pre-quantized weights for GPTQ and AWQ here: [CursorCore-Quantization](https://huggingface.co/collections/TechxGenus/cursorcore-quantization-67066431f29f252494ee8cf3)
|
| 82 |
|
| 83 |
+
We use the manually written benchmark APEval to assess the model's ability to assist programming. We also utilize [EvalPlus](https://github.com/evalplus/evalplus), [CanItEdit](https://github.com/nuprl/CanItEdit) and [OctoPack](https://github.com/bigcode-project/octopack) to evaluate the model's performance in Python program generation, instructional code editing, and automated program repair. Since we use a custom conversation template, its generation method differs significantly from both instruct models and base models. Please refer to [our paper](http://arxiv.org/abs/2410.07002) for more details.
|
| 84 |
+
|
| 85 |
+
Evaluation results on APEval:
|
| 86 |
+
|
| 87 |
+
<img src="https://github.com/TechxGenus/CursorCore/blob/main/pictures/APEval.png" alt="APEval" width="75%"/>
|
| 88 |
+
|
| 89 |
+
Evaluation results on EvalPlus, CanItEdit and OctoPack:
|
| 90 |
+
|
| 91 |
+
<img src="https://github.com/TechxGenus/CursorCore/blob/main/pictures/EvalPlus_CanItEdit_OctoPack.png" alt="EvalPlus_CanItEdit_OctoPack" width="75%">
|
| 92 |
+
|
| 93 |
## Usage
|
| 94 |
|
| 95 |
Here are some examples of how to use our model:
|
|
|
|
| 156 |
{
|
| 157 |
"type": "code",
|
| 158 |
"lang": "python",
|
| 159 |
+
"code": """def quick_sort(arr):
|
| 160 |
+
if len(arr) <= 1:
|
| 161 |
+
return arr
|
| 162 |
+
pivot = arr[len(arr) // 2]
|
| 163 |
+
left = [x for x in arr if x < pivot]
|
| 164 |
+
middle = [x for x in arr if x == pivot]
|
| 165 |
+
right = [x for x in arr if x > pivot]
|
| 166 |
+
return quick_sort(left) + middle + quick_sort(right)"""
|
| 167 |
}
|
| 168 |
],
|
| 169 |
"current": {
|
| 170 |
"type": "code",
|
| 171 |
"lang": "python",
|
| 172 |
+
"code": """def quick_sort(array):
|
| 173 |
+
if len(arr) <= 1:
|
| 174 |
+
return arr
|
| 175 |
+
pivot = arr[len(arr) // 2]
|
| 176 |
+
left = [x for x in arr if x < pivot]
|
| 177 |
+
middle = [x for x in arr if x == pivot]
|
| 178 |
+
right = [x for x in arr if x > pivot]
|
| 179 |
+
return quick_sort(left) + middle + quick_sort(right)"""
|
| 180 |
},
|
| 181 |
"user": ""
|
| 182 |
}
|
|
|
|
| 258 |
"current": {
|
| 259 |
"type": "code",
|
| 260 |
"lang": "python",
|
| 261 |
+
"code": """def quick_sort(array):
|
| 262 |
+
if len(arr) <= 1:
|
| 263 |
+
return arr
|
| 264 |
+
pivot = arr[len(arr) // 2]
|
| 265 |
+
left = [x for x in arr if x < pivot]
|
| 266 |
+
middle = [x for x in arr if x == pivot]
|
| 267 |
+
right = [x for x in arr if x > pivot]
|
| 268 |
+
return quick_sort(left) + middle + quick_sort(right)"""
|
| 269 |
},
|
| 270 |
"user": "Add Docstring."
|
| 271 |
}
|
|
|
|
| 336 |
"current": {
|
| 337 |
"type": "code",
|
| 338 |
"lang": "python",
|
| 339 |
+
"code": """def quick_sort(array):
|
| 340 |
+
if len(arr) <= 1:
|
| 341 |
+
return arr
|
| 342 |
+
pivot = arr[len(arr) // 2]
|
| 343 |
+
left = [x for x in arr if x < pivot]
|
| 344 |
+
middle = [x for x in arr if x == pivot]
|
| 345 |
+
right = [x for x in arr if x > pivot]
|
| 346 |
+
return quick_sort(left) + middle + quick_sort(right)"""
|
| 347 |
},
|
| 348 |
"user": "Add Docstring."
|
| 349 |
}
|
|
|
|
| 412 |
"current": {
|
| 413 |
"type": "code",
|
| 414 |
"lang": "python",
|
| 415 |
+
"code": """def quick_sort(array):
|
| 416 |
+
if len(arr) <= 1:
|
| 417 |
+
return arr
|
| 418 |
+
pivot = arr[len(arr) // 2]
|
| 419 |
+
left = [x for x in arr if x < pivot]
|
| 420 |
+
middle = [x for x in arr if x == pivot]
|
| 421 |
+
right = [x for x in arr if x > pivot]
|
| 422 |
+
return quick_sort(left) + middle + quick_sort(right)"""
|
| 423 |
},
|
| 424 |
"user": "Add Docstring."
|
| 425 |
}
|
|
|
|
| 489 |
}
|
| 490 |
```
|
| 491 |
|
| 492 |
+
## Acknowledgements
|
| 493 |
+
|
| 494 |
+
The open-source community has been of great help to us, and we reference numerous projects and applications. They include but are not limited to:
|
| 495 |
|
| 496 |
+
[Deepseek-Coder](https://github.com/deepseek-ai/DeepSeek-Coder), [Yi-Coder](https://github.com/01-ai/Yi-Coder), [Qwen-Coder](https://github.com/QwenLM/Qwen2.5-Coder), [Self-Instruct](https://github.com/yizhongw/self-instruct), [Evol-Instruct](https://github.com/theblackcat102/evol-dataset), [OSS-Instruct](https://github.com/ise-uiuc/magicoder), [EvalPlus](https://github.com/evalplus/evalplus), [CanItEdit](https://github.com/nuprl/CanItEdit), [OctoPack](https://github.com/bigcode-project/octopack), [Aider](https://github.com/Aider-AI/aider), [Continue](https://github.com/continuedev/continue), [Cursor](https://github.com/getcursor/cursor), ...
|
| 497 |
|
| 498 |
+
## Contribution
|
| 499 |
|
| 500 |
+
Contributions are welcome! If you find any bugs or have suggestions for improvements, please open an issue or submit a pull request.
|