Update README.md

README.md

@@ -63,20 +63,20 @@ This gives LED excellent performance on structural legal documents.
 | Setting | Value |
 |--------|--------|
 | Base model | nsi319/legal-led-base-16384 |
-| Epochs | 4 total |
-| Batch size | 1 |
-| Gradient accumulation | 8 |
+| Epochs | 6 total |
+| Batch size | 2 |
+| Gradient accumulation | 2 |
 | Learning rate | 1e-5 |
 | Optimizer | AdamW |
 | Weight decay | 0.01 |
 | FP16 | Yes |
 | Warmup steps | 500 |
-| Max input length | 14,000 tokens |
+| Max input length | 4096 tokens |
 | Max output length | 512 tokens |
 | Attention | Global attention on first token |
 | Scheduler | Linear |
 
-Training was performed on **NVIDIA T4 (16GB VRAM)** via Kaggle.
+Training was performed on **NVIDIA P100 (16GB VRAM)** via Kaggle.
 
 ---