Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -19,12 +19,26 @@ size_categories:
|
|
19 |
|
20 |
a "stacked" version of `xsum`
|
21 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
## updates
|
23 |
|
24 |
- dec 3: upload initial version
|
25 |
- dec 4: upload v2 with basic data quality fixes (i.e. the `is_stacked` column)
|
26 |
- dec 5 0500: upload v3 which has pre-randomised order and duplicate rows for document+summary dropped
|
27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
## dataset details
|
29 |
|
30 |
see the repo `.log` file for more details.
|
|
|
19 |
|
20 |
a "stacked" version of `xsum`
|
21 |
|
22 |
+
1. Original Dataset: copy of the base dataset
|
23 |
+
|
24 |
+
2. Stacked Rows: The original dataset is processed by stacking rows based on certain criteria:
|
25 |
+
- Maximum Input Length: The maximum length for input sequences is 1024 tokens in the longt5 model tokenizer.
|
26 |
+
- Maximum Output Length: The maximum length for output sequences is also 1024 tokens in the longt5 model tokenizer.
|
27 |
+
|
28 |
+
3. Special Token: The dataset utilizes the `[NEXT_CONCEPT]` token to indicate a new topic **within** the same summary. It is recommended to explicitly add this special token to your model's tokenizer before training, ensuring that it is recognized and processed correctly during downstream usage.
|
29 |
+
4.
|
30 |
## updates
|
31 |
|
32 |
- dec 3: upload initial version
|
33 |
- dec 4: upload v2 with basic data quality fixes (i.e. the `is_stacked` column)
|
34 |
- dec 5 0500: upload v3 which has pre-randomised order and duplicate rows for document+summary dropped
|
35 |
|
36 |
+
|
37 |
+
## stats
|
38 |
+
|
39 |
+
|
40 |
+
![stats](https://i.imgur.com/TyyDthT.png)
|
41 |
+
|
42 |
## dataset details
|
43 |
|
44 |
see the repo `.log` file for more details.
|