Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,24 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
This is full finetune model from RWKV 4 world 7B CHNTuned using data from Readflow tech (readflow.com.cn) ,
|
| 6 |
+
|
| 7 |
+
finetuned for 32k context length, used to summary news article
|
| 8 |
+
|
| 9 |
+
using inf-ctx training https://github.com/SynthiaDL/TrainChatGalRWKV
|
| 10 |
+
|
| 11 |
+
you can test summary prompt using RWKV runner in chat mode , and check conversation files in examples folders.
|
| 12 |
+
|
| 13 |
+
https://discord.gg/pWH5MkvtNR
|
| 14 |
+
|
| 15 |
+
-------------------------------------------------------------------------------------------------
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
这是和会读Readflow合作的模型（ readflow.com.cn），用于超长微信文章的摘要，训练了32k的context长度，能一次性输入整篇文章进行摘要，如下图以及在example文件夹中的例子，23k tokens一次性输入摘要，
|
| 19 |
+
|
| 20 |
+
RWKV world的token对中文以及各类语言效率很高，token和字比例，基本是1:1,甚至1:n，欢迎测试，并加入微调QQ群讨论439087067
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
|
| 24 |
+
|