[BREAKING][reward] refactor: deprecate batch reward manager (verl-project#5237)

### What does this PR do?

This PR does the following refactor effort:

1. Unify the reward managers (i.e., `config.reward_manager` and
`config.reward_model.reward_manager`) into one, accepting both the
`register` reward manager and the user-customized module.
2. Remove legacy implementation and modify the relevant scripts and
docs.

```
reward_manager:
  _target_: verl.workers.config.reward_model.RewardManagerConfig
  source: register
  name: naive
  module:
    _target_: verl.trainer.config.config.ModuleConfig
    path: null
    name: custom_reward_manager
```

### Checklist Before Starting

- [ ] Search for similar PRs. Paste at least one query link here: ...
- [ ] Format the PR title as `[{modules}] {type}: {description}` (This
will be checked by the CI)
- `{modules}` include `fsdp`, `megatron`, `veomni`, `sglang`, `vllm`,
`rollout`, `trainer`, `ci`, `training_utils`, `recipe`, `hardware`,
`deployment`, `ray`, `worker`, `single_controller`, `misc`, `perf`,
`model`, `algo`, `env`, `tool`, `ckpt`, `doc`, `data`, `cfg`, `reward`
- If this PR involves multiple modules, separate them with `,` like
`[megatron, fsdp, doc]`
  - `{type}` is in `feat`, `fix`, `refactor`, `chore`, `test`
- If this PR breaks any API (CLI arguments, config, function signature,
etc.), add `[BREAKING]` to the beginning of the title.
  - Example: `[BREAKING][fsdp, megatron] feat: dynamic batching`

### Test

> For changes that can not be tested by CI (e.g., algorithm
implementation, new model support), validate by experiment(s) and show
results like training curve plots, evaluation results, etc.

### API and Usage Example

> Demonstrate how the API changes if any, and provide usage example(s)
if possible.

```python
# Add code snippet or script demonstrating how to use this
```

### Design & Code Changes

> Demonstrate the high-level design if this PR is complex, and list the
specific changes.

### Checklist Before Submitting

> [!IMPORTANT]
> Please check all the following items before requesting a review,
otherwise the reviewer might deprioritize this PR for review.

- [ ] Read the [Contribute
Guide](https://github.com/volcengine/verl/blob/main/CONTRIBUTING.md).
- [ ] Apply [pre-commit
checks](https://github.com/volcengine/verl/blob/main/CONTRIBUTING.md#code-linting-and-formatting):
`pre-commit install && pre-commit run --all-files --show-diff-on-failure
--color=always`
- [ ] Add / Update [the
documentation](https://github.com/volcengine/verl/tree/main/docs).
- [ ] Add unit or end-to-end test(s) to [the CI
workflow](https://github.com/volcengine/verl/tree/main/.github/workflows)
to cover all the code. If not feasible, explain why: ...
- [ ] Once your PR is ready for CI, send a message in [the `ci-request`
channel](https://verl-project.slack.com/archives/C091TCESWB1) in [the
`verl` Slack
workspace](https://join.slack.com/t/verl-project/shared_invite/zt-3855yhg8g-CTkqXu~hKojPCmo7k_yXTQ).
(If not accessible, please try [the Feishu group
(飞书群)](https://applink.larkoffice.com/client/chat/chatter/add_by_link?link_token=772jd4f1-cd91-441e-a820-498c6614126a).)
- [ ] If your PR is related to the `recipe` submodule, please also
update the reference to the submodule commit via `git submodule update
--remote` or `cd recipe && git pull origin main`.

Author: yyDing1

docs/advance/reward_loop.rst

@@ -204,7 +204,7 @@ See ``verl/experimental/reward_manager/*`` for reference.
          # your own reward manager
          ...
 
-After defining it, users can specify their custom reward manager by setting ``reward_model.reward_manager=user_costomized``.
+After defining it, users can specify their custom reward manager by setting ``reward_model.reward_manager.name=user_costomized``.
 
 RewardLoopManager
 ~~~~~~~~~~~~~~~~~

docs/ascend_tutorial/examples/dapo_multi_model_optimization_practice.md

@@ -15,7 +15,7 @@ DAPO的论文可以参考：[DAPO](https://arxiv.org/pdf/2503.14476)，其中包
   在dapo算法中，必须配置成dapo。
 
 ```
-reward_model.reward_manager=dapo
+reward_model.reward_manager.name=dapo
 ```
 
 - **Clip-Higher 更高裁剪 **
@@ -250,7 +250,7 @@ ray job submit --no-wait --runtime-env="${RUNTIME_ENV}" \
     actor_rollout_ref.ref.fsdp_config.param_offload=True \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=${fsdp_size} \
-    reward_model.reward_manager=dapo \
+    reward_model.reward_manager.name=dapo \
     reward_model.overlong_buffer.enable=${enable_overlong_buffer} \
     reward_model.overlong_buffer.len=${overlong_buffer_len} \
     reward_model.overlong_buffer.penalty_factor=${overlong_penalty_factor} \

examples/gmpo_trainer/test_dapo_7b_math.sh

@@ -117,7 +117,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=${fsdp_size} \
     actor_rollout_ref.actor.checkpoint.save_contents="${save_contents}" \
-    reward_model.reward_manager=dapo \
+    reward_model.reward_manager.name=dapo \
     +reward_model.reward_kwargs.overlong_buffer_cfg.enable=${enable_overlong_buffer} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.len=${overlong_buffer_len} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.penalty_factor=${overlong_penalty_factor} \

examples/gmpo_trainer/test_dapo_qwen3_30b_math.sh

@@ -113,7 +113,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=${fsdp_size} \
     actor_rollout_ref.actor.checkpoint.save_contents="${save_contents}" \
-    reward_model.reward_manager=dapo \
+    reward_model.reward_manager.name=dapo \
     +reward_model.reward_kwargs.overlong_buffer_cfg.enable=${enable_overlong_buffer} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.len=${overlong_buffer_len} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.penalty_factor=${overlong_penalty_factor} \

examples/grpo_trainer/run_deepseek671b_math_megatron_96gb.sh

@@ -159,7 +159,7 @@ python3 -m verl.trainer.main_ppo \
     +actor_rollout_ref.actor.megatron.override_transformer_config.account_for_embedding_in_pipeline_split=False \
     +actor_rollout_ref.actor.megatron.override_transformer_config.account_for_loss_in_pipeline_split=False \
     +actor_rollout_ref.actor.megatron.override_transformer_config.num_layers_in_last_pipeline_stage=${LAST_LAYER} \
-    reward_model.reward_manager=dapo \
+    reward_model.reward_manager.name=dapo \
     +reward_model.reward_kwargs.overlong_buffer_cfg.enable=${enable_overlong_buffer} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.len=${overlong_buffer_len} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.penalty_factor=${overlong_penalty_factor} \

examples/grpo_trainer/run_qwen3-235b_megatron_96gb.sh

@@ -161,7 +161,7 @@ python3 -m verl.trainer.main_ppo \
     +actor_rollout_ref.actor.megatron.override_transformer_config.moe_enable_deepep=True \
     +actor_rollout_ref.actor.megatron.override_transformer_config.account_for_loss_in_pipeline_split=True \
     +actor_rollout_ref.actor.megatron.override_transformer_config.account_for_embedding_in_pipeline_split=True \
-    reward_model.reward_manager=dapo \
+    reward_model.reward_manager.name=dapo \
     +reward_model.reward_kwargs.overlong_buffer_cfg.enable=${enable_overlong_buffer} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.len=${overlong_buffer_len} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.penalty_factor=${overlong_penalty_factor} \

examples/grpo_trainer/run_qwen3moe-30b_megatron_96gb.sh

@@ -176,7 +176,7 @@ python3 -m verl.trainer.main_ppo --config-path=./config --config-name='ppo_megat
     actor_rollout_ref.ref.megatron.context_parallel_size=${REF_CP} \
     actor_rollout_ref.ref.megatron.expert_model_parallel_size=${REF_EP} \
     actor_rollout_ref.ref.megatron.expert_tensor_parallel_size=${REF_ETP} \
-    reward_model.reward_manager=dapo \
+    reward_model.reward_manager.name=dapo \
     +reward_model.reward_kwargs.overlong_buffer_cfg.enable=${enable_overlong_buffer} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.len=${overlong_buffer_len} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.penalty_factor=${overlong_penalty_factor} \

examples/gspo_trainer/run_qwen30b_gspo.sh

@@ -153,7 +153,7 @@ ROLLOUT_CONFIG="
 
 # ===================================== Reward =====================================
 REWARD_CONFIG="
-    reward_model.reward_manager=dapo \
+    reward_model.reward_manager.name=dapo \
     +reward_model.reward_kwargs.overlong_buffer_cfg.enable=${enable_overlong_buffer} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.len=${overlong_buffer_len} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.penalty_factor=${overlong_penalty_factor} \

examples/gspo_trainer/test_gspo_3b_math.sh

@@ -173,7 +173,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.ref.fsdp_config.param_offload=${offload} \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.entropy_checkpointing=${entropy_checkpointing} \
-    reward_model.reward_manager=${reward_manager} \
+    reward_model.reward_manager.name=${reward_manager} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.enable=${enable_overlong_buffer} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.len=${overlong_buffer_len} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.penalty_factor=${overlong_penalty_factor} \

examples/gspo_trainer/test_gspo_3b_math_slurm.sh

@@ -177,7 +177,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.ref.fsdp_config.param_offload=${offload} \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.entropy_checkpointing=${entropy_checkpointing} \
-    reward_model.reward_manager=${reward_manager} \
+    reward_model.reward_manager.name=${reward_manager} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.enable=${enable_overlong_buffer} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.len=${overlong_buffer_len} \
     +reward_model.reward_kwargs.overlong_buffer_cfg.penalty_factor=${overlong_penalty_factor} \