refactor: 训练循环改为两重迭代并统一参数命名

- 训练循环从三重(epoch→batched→batch)改为二重(epoch→batch)
- batch_size → batch_per_device, accumulation_steps → grad_accum_steps
- scheduler 移入 step block 对齐 optimizer 更新步
- GradientClippingCallback 改用 on_step_begin 避免零梯度裁剪
- 移除 _train_impl 误导性的 -> Checkpoint 标注
- total_steps 修除为向下取整并精简为一行
- warmup_steps 改为 warmup_ratio (默认0.05)

README.md

@@ -78,15 +78,27 @@ Or download manually from [HuggingFace](https://huggingface.co/ViperEk/KHAOSZ) i
 #### Train a Model
 
 ```bash
-CUDA_VISIBLE_DEVICES=0,1,2,3 python scripts/tools/train.py \
-    --train_type seq \
-    --data_root_path /path/to/dataset \
-    --param_path /path/to/model \
-    --batch_size 4 \
-    --accumulation_steps 8 \
-    --max_lr 3e-4 \
-    --warmup_steps 1000 \
-    --n_epoch 1
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+
+nohup python scripts/tools/train.py \
+    --nprocs=4 \
+    --train_type=sft \
+    --data_root_path=/path/to/dataset \
+    --param_path=/path/to/model \
+    --batch_per_device=4 \
+    --grad_accum_steps=8 \
+    --warmup_ratio=0.05 \
+    --max_lr=1e-4 \
+    --max_grad_norm=1.0 \
+    --adamw_beta1=0.99 \
+    --adamw_beta2=0.95 \
+    --adamw_weight_decay=1e-5 \
+    --window_size=2048 \
+    --ckpt_interval=10000 \
+    --ckpt_dir=./checkpoint \
+    --random_seed=3407 \
+    --label_smoothing=0.1 \
+    > out.log 2> err.log &
 ```
 
 Full reference at [Parameter Guide](assets/docs/params.md).