Update README.md
Browse files
README.md
CHANGED
|
@@ -46,7 +46,7 @@ Wikipedia simple data splitted using Langchain's RecursiveCharacterTextSplitter
|
|
| 46 |
- This can be used for Annealing LLMs.
|
| 47 |
## Why its different
|
| 48 |
|
| 49 |
-
- This dataset is split with a max length of 448 (128*3.5) characters
|
| 50 |
- And rather than splitting by length, it is split using RecursiveCharacterTextSplitter. So, the chunks don't end randomly.
|
| 51 |
- Can use very large batch sizes.
|
| 52 |
|
|
|
|
| 46 |
- This can be used for Annealing LLMs.
|
| 47 |
## Why its different
|
| 48 |
|
| 49 |
+
- This dataset is split with a max length of 448 (128*3.5) and 1024 (256*4) characters
|
| 50 |
- And rather than splitting by length, it is split using RecursiveCharacterTextSplitter. So, the chunks don't end randomly.
|
| 51 |
- Can use very large batch sizes.
|
| 52 |
|