Update README.md
Browse files
README.md
CHANGED
@@ -6,18 +6,18 @@ language:
|
|
6 |
- en
|
7 |
license: cc-by-sa-4.0
|
8 |
---
|
9 |
-
|
10 |
-
|
11 |
|
12 |
-
|
13 |
-
1. The data format is converted from "chatml" to "messages", which is more convenient to use ```tokenizer.apply_chat_template```.
|
14 |
-
2.
|
15 |
-
3. The old version has been moved to "legacy" branch.
|
16 |
|
|
|
|
|
|
|
17 |
* [Paper](https://arxiv.org/abs/2312.11193)
|
18 |
* [Github](https://github.com/yuyijiong/train_with_paraphrasing)
|
19 |
-
|
20 |
-
# Long Context Instruction-tuning dataset with "Original text paraphrasing"
|
21 |
* consist of multiple tasks
|
22 |
* Chinese and English
|
23 |
* sample length ranging from 4k to 32k
|
|
|
6 |
- en
|
7 |
license: cc-by-sa-4.0
|
8 |
---
|
9 |
+
## 🔥 Updates
|
10 |
+
**\[2024.6.4\]** Add a slim version. The sample number is reduced from about 20k to 10k.
|
11 |
|
12 |
+
**\[2024.5.28\]**
|
13 |
+
1. The data format is converted from "chatml" to "messages", which is more convenient to use ```tokenizer.apply_chat_template```. The old version has been moved to "legacy" branch.
|
14 |
+
2. The version without "Original text paraphrasing" is added.
|
|
|
15 |
|
16 |
+
|
17 |
+
|
18 |
+
# 📊 Long Context Instruction-tuning dataset with "Original text paraphrasing"
|
19 |
* [Paper](https://arxiv.org/abs/2312.11193)
|
20 |
* [Github](https://github.com/yuyijiong/train_with_paraphrasing)
|
|
|
|
|
21 |
* consist of multiple tasks
|
22 |
* Chinese and English
|
23 |
* sample length ranging from 4k to 32k
|