Skip to content

Commit

Permalink
Update README.md
Browse files Browse the repository at this point in the history
  • Loading branch information
Y1ran authored Sep 29, 2020
1 parent f8bc54b commit 888af6a
Showing 1 changed file with 5 additions and 5 deletions.
10 changes: 5 additions & 5 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@ EssayKiller是基于OCR、NLP领域的最新模型所构建的生成式文本创

| 项目作者 | 主页1 | 主页2 | 主页3 |
| ------------- |:-------------:|:----:|:---:|
| 图灵的猫 | [知乎](https://www.zhihu.com/people/dong-xi-97-29) | [B站](https://space.bilibili.com/371846699) | Youtube |
| 图灵的猫 | [知乎](https://www.zhihu.com/people/dong-xi-97-29) | [B站](https://space.bilibili.com/371846699) | [Youtube](https://www.youtube.com/channel/UCoEVP6iTw5sfozUGLLWJyDg/channels) |


**致谢**
Expand Down Expand Up @@ -227,7 +227,7 @@ python run.py --model bert
**预训练**
参考 [GPT2-ML](https://github.com/imcaspar/gpt2-ml/) 预训练模型,使用 [Quadro RTX 8000](https://www.nvidia.com/en-us/design-visualization/quadro/rtx-8000/) 训练 28w 步

>![](https://github.com/EssayKillerBrain/EssayKiller_V2/blob/master/References/attachments/2233.png)
>![](https://github.com/EssayKillerBrain/EssayKiller_V2/blob/master/References/attachments/2233.PNG)

<br>
Expand Down Expand Up @@ -299,10 +299,10 @@ CUDA_VISIBLE_DEVICES=0 python train/train_wc.py --input_file=/data/EssayKiller/

| 模型 | 参数量 | 下载链接 | 备注 |
| ------------- |:-------------:|:----:|:---:|
| EAST | < 0.1 Billion | [网盘链接](https://pan.baidu.com/s/1S9WpyXZ_wVWbKo-WsB0q5A) 提取码:sxhh| 检测模型 |
| EAST | < 0.1 Billion | [GoogleDrive](https://pan.baidu.com/s/1S9WpyXZ_wVWbKo-WsB0q5A) 提取码:sxhh| 检测模型 |
| CRNN | < 0.1 Billion | [网盘链接](https://eyun.baidu.com/s/3dEUJJg9) 提取码:vKeD| 识别模型 |
| BERT | 0.1 Billion | [网盘链接](https://pan.baidu.com/s/1BrdFSx9_n1q2uWBiQrpalw) 提取码:kim2| 摘要模型 |
| GPT-2 | 1.5 Billion | | 生成模型 |
| BERT | 0.1 Billion | [GoogleDrive](https://pan.baidu.com/s/1BrdFSx9_n1q2uWBiQrpalw) 提取码:kim2| 摘要模型 |
| GPT-2 | 1.5 Billion | [GoogleDrive](https://drive.google.com/file/d/1ujWYTOvRLGJX0raH-f-lPZa3-RN58ZQx/view?usp=sharing) | 生成模型 |

整个AI的参数量分布不均匀,主要原因在于,这是一个语言类AI,99%的参数量集中在语言网络中,其中GPT-2(15亿)占88%,BERT(1.1亿)占7%,其他的识别网络和判分网络共占5%。

Expand Down

0 comments on commit 888af6a

Please sign in to comment.