Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets:
|
| 3 |
+
- hlillemark/c4_t5_corrupted_seqlen256
|
| 4 |
+
language:
|
| 5 |
+
- en
|
| 6 |
+
metrics:
|
| 7 |
+
- perplexity
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
|Hyperparameter |Value |
|
| 11 |
+
|---------------------|---------|
|
| 12 |
+
|Steps | 150k|
|
| 13 |
+
|Max length | 256|
|
| 14 |
+
|LR | 1e-4|
|
| 15 |
+
|LR schedule | constant|
|
| 16 |
+
|Optimizer | AdamW|
|
| 17 |
+
|beta_1, beta_2 |0.9, 0.95|
|
| 18 |
+
|Final eval loss | 2.245|
|
| 19 |
+
|Final eval perplexity| 9.44|
|