FarisHijazi
commited on
Commit
•
1b08789
1
Parent(s):
6e378ca
Update README.md
Browse files
README.md
CHANGED
@@ -18,7 +18,7 @@ There are many public datasets in Kajiwoto, the power in this dataset is the met
|
|
18 |
|
19 |
Do be aware that a lot of the data is NSFW (explicit content)
|
20 |
|
21 |
-
The raw datasets are in [
|
22 |
1. transform shape (convert to a known format such as ShareGPT)
|
23 |
2. deduplication
|
24 |
3. template rendering of strings such as `"you rolled a dice with %{1|2|3|4|5|6}"`. This operation is lossy as it will choose only one of the options
|
@@ -26,7 +26,7 @@ The raw datasets are in [./kajiwoto_raw.json](), this data needs to be processed
|
|
26 |
5. dropping datasets with too few upvotes or comments
|
27 |
6. filtering in or out NSFW datasets
|
28 |
|
29 |
-
I have processed an initial example here: [
|
30 |
|
31 |
it is any dataset with at least 1 upvote and at least 6 lines in the conversation, you can most models as this is in the shareGPT format
|
32 |
|
|
|
18 |
|
19 |
Do be aware that a lot of the data is NSFW (explicit content)
|
20 |
|
21 |
+
The raw datasets are in [kajiwoto_raw.json](./kajiwoto_raw.json), this data needs to be processed so that it can be used, the main operations are:
|
22 |
1. transform shape (convert to a known format such as ShareGPT)
|
23 |
2. deduplication
|
24 |
3. template rendering of strings such as `"you rolled a dice with %{1|2|3|4|5|6}"`. This operation is lossy as it will choose only one of the options
|
|
|
26 |
5. dropping datasets with too few upvotes or comments
|
27 |
6. filtering in or out NSFW datasets
|
28 |
|
29 |
+
I have processed an initial example here: [kajiwoto_sharegpt-len_gt_6-upvotes_gt_0-sampled.json](./kajiwoto_sharegpt-len_gt_6-upvotes_gt_0-sampled.json)
|
30 |
|
31 |
it is any dataset with at least 1 upvote and at least 6 lines in the conversation, you can most models as this is in the shareGPT format
|
32 |
|