modified model card

README.md

@@ -25,15 +25,14 @@ Training was conducted on a single NVIDIA DGX node:
 
 - Hardware: **8× NVIDIA A100 GPUs (80GB HBM each)**
 - Training duration: **1440 GPU hours** (~180 hours wall-clock on 8 GPUs)
-- Total epochs: **5 epochs**
+- Total steps: **160,000 global steps**
 
 ### 📈 Training Loss Curve
 
 Here's the training loss progression:
 
-```markdown
 ![Training Loss Curve](../plots/pretrain_loss_20250303.png)
-```
+
 
 ### 📝 Example Outputs
 Below are generated examples illustrating Argonne-1.0's style and capability when prompted: