Update README.md
Browse files
README.md
CHANGED
|
@@ -40,7 +40,7 @@ All translated (if not with DeepL) with a combination of GeminiFlash 1.5/GPT4o m
|
|
| 40 |
* Learning rate: 2e-4
|
| 41 |
* Weight decay: 1e-3
|
| 42 |
* Learning schedule: linear, with 5_000 warmup steps
|
| 43 |
-
* Num epochs: ~3
|
| 44 |
|
| 45 |
Train perplexity: 2.5
|
| 46 |
Validation perplexity: 3.4
|
|
|
|
| 40 |
* Learning rate: 2e-4
|
| 41 |
* Weight decay: 1e-3
|
| 42 |
* Learning schedule: linear, with 5_000 warmup steps
|
| 43 |
+
* Num epochs: ~3 (off-premise) followed by 3 (on-premise)
|
| 44 |
|
| 45 |
Train perplexity: 2.5
|
| 46 |
Validation perplexity: 3.4
|