Commit
·
876cc8e
1
Parent(s):
bf96279
Update README.md
Browse files
README.md
CHANGED
|
@@ -12,3 +12,19 @@ inference:
|
|
| 12 |
min_length: 128
|
| 13 |
max_length: 512
|
| 14 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
min_length: 128
|
| 13 |
max_length: 512
|
| 14 |
---
|
| 15 |
+
|
| 16 |
+
This is the model trained for this video:
|
| 17 |
+
|
| 18 |
+
https://www.youtube.com/watch?v=OEPL5Tm3mmQ
|
| 19 |
+
|
| 20 |
+
Due to hardware limitations, I trained this model with only a batch size of 2. (I know this isn't ideal).
|
| 21 |
+
|
| 22 |
+
The quality of the model may be affected.
|
| 23 |
+
|
| 24 |
+
After training was complete, the best model according to a hold-out set was used.
|
| 25 |
+
|
| 26 |
+
This model was trained using a filtered version of this dataset:
|
| 27 |
+
|
| 28 |
+
https://www.kaggle.com/datasets/thomaskonstantin/3500-popular-creepypastas
|
| 29 |
+
|
| 30 |
+
This dataset had a lot of blank entries and missing text.
|