Fix default values in Key arguments table

Author: delock

training/deepspeed_finetune_demo/README.md

@@ -20,14 +20,14 @@ For example, if we want to run Qwen2.5-3B model with ZeRO offload on 2 GPUs, we
 | Argument | Description | Default |
 |----------|-------------|---------|
 | `--batch_size` | Training batch size per GPU | required |
-| `--eval_batch_size` | Eval batch size per rank | 1 |
+| `--eval_batch_size` | Eval batch size per rank | 4 |
 | `--eval_steps` | Run evaluation every N steps (0 disables) | 0 |
 | `--max_steps` | Stop after N steps (-1 = full epoch) | -1 |
 | `--checkpoint_steps` | Save a checkpoint every N steps (0 disables); keeps last 2 | 0 |
 | `--wandb_name` | Wandb run name (optional) | None |
-| `--num_train_epochs` | Number of training epochs | 1 |
+| `--num_train_epochs` | Number of training epochs | 3 |
 | `--weight_decay` | Weight decay | 0.01 |
-| `--warmup` | Warmup steps | 0 |
+| `--warmup` | Warmup ratio | 0.01 |
 
 Note: Learning rate is controlled entirely by the DeepSpeed config JSON, not by command-line arguments.