Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
tags:
|
| 5 |
+
- question-answering
|
| 6 |
+
license: mit
|
| 7 |
+
datasets:
|
| 8 |
+
- SQuAD
|
| 9 |
+
metrics:
|
| 10 |
+
- EM
|
| 11 |
+
- F1
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# Test model for DL4NLP 2022 HW06
|
| 15 |
+
xtremedistil-l6-h256-uncased trained on SQuAD
|
| 16 |
+
|
| 17 |
+
## Hyper parameters
|
| 18 |
+
- learning rate: 1e-5
|
| 19 |
+
- weight decay: 0.01
|
| 20 |
+
- warm up steps: 0
|
| 21 |
+
- learning rate scheduler: linear
|
| 22 |
+
- epochs: 1
|
| 23 |
+
|
| 24 |
+
## Metric results on the dev set
|
| 25 |
+
- F1: 65.48
|
| 26 |
+
- EM: 51.67
|