fix: 统一 OPD 训练类型值

- 将 SDK 请求和示例中的训练类型调整为 pg_opd
- 更新中英文文档与 OPD 单元测试

Author: agentscope-ai-java

dashscope/finetune/reinforcement/common/model_types.py

@@ -47,7 +47,7 @@ def __str__(self):
 
 class TrainingType(str, Enum):
     TRAINING_TYPE = "reinforcement"
-    PG_OPD = "pg-opd"
+    PG_OPD = "pg_opd"
 
     def __str__(self):
         return self.value

dashscope/finetune/reinforcement/examples/workspace/UserGuide-zh.md

@@ -345,7 +345,7 @@ functions:
   # ...
 
 training:
-  type: pg-opd
+  type: pg_opd
 ```
 
 ```bash

dashscope/finetune/reinforcement/examples/workspace/UserGuide.md

@@ -334,7 +334,7 @@ functions:
   # ...
 
 training:
-  type: pg-opd
+  type: pg_opd
 ```
 
 ```bash

dashscope/finetune/reinforcement/examples/workspace/opd-job.yaml

@@ -58,7 +58,7 @@ datasets:
   data_source_type: file_id
   file_name: ./data/calc_validation_min.jsonl
 training:
-  type: pg-opd
+  type: pg_opd
   hyper_parameters:
     eval_steps: 1
     kl_loss_coef: 0.002

tests/unit/test_opd.py

@@ -42,7 +42,7 @@ def _success_response():
 
 
 def test_pg_opd_training_type_is_public():
-    assert TrainingType.PG_OPD == "pg-opd"
+    assert TrainingType.PG_OPD == "pg_opd"
 
 
 @pytest.mark.parametrize(