Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,28 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- Skylion007/openwebtext
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
pipeline_tag: text-generation
|
| 8 |
+
---
|
| 9 |
+
### gpt trained with nanoGPT
|
| 10 |
+
|
| 11 |
+
Configs:
|
| 12 |
+
- batch size = 32
|
| 13 |
+
- bias = False
|
| 14 |
+
- bloack_size = 1024
|
| 15 |
+
- n heads = 8
|
| 16 |
+
- h layers = 6
|
| 17 |
+
- dropout = 0.0
|
| 18 |
+
- n embed = 768
|
| 19 |
+
- vocab size = 50304
|
| 20 |
+
- gradient_accumulation_steps = 1
|
| 21 |
+
- learning_rate = 1e-3
|
| 22 |
+
- iters = 7250
|
| 23 |
+
- lr_decay_iters = 5000
|
| 24 |
+
- min_lr = 1e-5
|
| 25 |
+
- warmup_iters = 400
|
| 26 |
+
- mfu = 30.45935
|
| 27 |
+
- train_loss = 3.89759
|
| 28 |
+
- val_loss = 3.91001
|