Commit
•
57d2403
1
Parent(s):
780b2f9
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,7 +1,23 @@
|
|
| 1 |
-
---
|
| 2 |
datasets:
|
| 3 |
-
- roneneldan/TinyStories
|
| 4 |
metrics:
|
| 5 |
-
- glue
|
| 6 |
-
- super_glue
|
| 7 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
datasets:
|
| 2 |
+
- roneneldan/TinyStories
|
| 3 |
metrics:
|
| 4 |
+
- glue
|
| 5 |
+
- super_glue
|
| 6 |
+
---
|
| 7 |
+
Basemodel: GPT-Neo
|
| 8 |
+
Configs:
|
| 9 |
+
Vocab size: 10,000
|
| 10 |
+
Hidden size: 512
|
| 11 |
+
Max position embeddings: 512
|
| 12 |
+
Number of layers: 2
|
| 13 |
+
Number of heads: 4
|
| 14 |
+
Window size: 256
|
| 15 |
+
Intermediate-size: 1024
|
| 16 |
+
|
| 17 |
+
Results:
|
| 18 |
+
- Task: glue
|
| 19 |
+
Score: 58.36
|
| 20 |
+
Confidence Interval: [57.95, 58.78]
|
| 21 |
+
- Task: blimp
|
| 22 |
+
Score: 55.64
|
| 23 |
+
Confidence Interval: [54.68, 56.64]
|