feat: GRPO CLI 接入 + on-policy，OpenAI API top_k 参数化，补充训练参数表

- train.py 新增 --train_type=grpo 及参数 (--grpo_clip_eps, --grpo_kl_coef, --group_size, --grpo_sync_interval, --start_epoch)
- GRPOStrategy 统一 on-policy 模式，ratio = exp(logπ_θ - logπ_ref)，PPO 裁剪目标，sync_interval 自动同步 ref_model
- ChatCompletionRequest 新增 top_k 参数，不再硬编码
- 补充 README 完整训练参数表（含此前缺失的 max_grad_norm / adamw / window_size / stride 等）

README.md

@@ -73,18 +73,34 @@ python scripts/tools/train.py --train_type=seq --data_root_path=/path/to/dataset
 
 | Parameter | Description | Default |
 |-----------|-------------|---------|
-| `--train_type` | Training type (`seq`, `sft`, `dpo`) | required |
+| `--train_type` | Training type (`seq`, `sft`, `dpo`, `grpo`) | required |
 | `--data_root_path` | Dataset root directory | required |
 | `--param_path` | Model / checkpoint path | required |
 | `--n_epoch` | Training epochs | 1 |
 | `--batch_size` | Batch size | 1 |
 | `--accumulation_steps` | Gradient accumulation steps | 1 |
-| `--max_lr` | Peak learning rate (cosine decay) | 3e-4 |
 | `--warmup_steps` | LR warmup steps | 1000 |
+| `--max_lr` | Peak learning rate (cosine decay) | 3e-4 |
+| `--max_grad_norm` | Max gradient norm for clipping | 1.0 |
+| `--adamw_beta1` | AdamW beta1 | 0.9 |
+| `--adamw_beta2` | AdamW beta2 | 0.95 |
+| `--adamw_weight_decay` | AdamW weight decay | 0.01 |
+| `--random_seed` | Random seed | 3407 |
+| `--num_workers` | DataLoader workers | 4 |
+| `--window_size` | Max input sequence length | auto |
+| `--stride` | Sequence stride | auto |
+| `--label_smoothing` | Label smoothing for cross entropy | 0.1 |
+| `--dpo_beta` | DPO beta | 0.1 |
+| `--grpo_clip_eps` | GRPO clip epsilon | 0.2 |
+| `--grpo_kl_coef` | GRPO KL penalty coefficient | 0.01 |
+| `--group_size` | GRPO group size | 4 |
+| `--grpo_sync_interval` | GRPO ref model sync interval (steps) | 200 |
 | `--ckpt_interval` | Checkpoint interval (iters) | 5000 |
 | `--ckpt_dir` | Checkpoint directory | checkpoint |
-| `--num_workers` | DataLoader workers | 4 |
+| `--start_epoch` | Start epoch (for resume) | 0 |
+| `--start_batch` | Start batch (for resume) | 0 |
 | `--nprocs` | Number of GPUs | 1 |
+| `--device_type` | Device type | cuda |
 
 Full reference at [Parameter Guide](./assets/docs/params.md#training-parameters).