[BREAKING] [rollout] chore: remove default rollout selection (#2757)

### What does this PR do?

As title

### 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`, `sglang`, `vllm`, `rollout`,
`trainer`, `ci`, `training_utils`, `recipe`, `hardware`, `deployment`,
`ray`, `worker`, `single_controller`, `misc`, `perf`, `model`, `algo`,
`env`, `tool`, `ckpt`, `doc`, `data`
- 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).)

examples/grpo_trainer/run_qwen3-236b_megatron.sh

@@ -111,6 +111,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.megatron.pipeline_model_parallel_size=${train_pp} \
     actor_rollout_ref.ref.megatron.tensor_model_parallel_size=${train_tp} \
     actor_rollout_ref.ref.megatron.expert_model_parallel_size=${train_ep} \

recipe/dapo/run_dapo_early_qwen2.5_32b.sh

@@ -108,6 +108,7 @@ ray job submit --no-wait --runtime-env="${RUNTIME_ENV}" \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.fsdp_config.param_offload=${offload} \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=-1 \

recipe/dapo/run_dapo_qwen2.5_32b.sh

@@ -110,6 +110,7 @@ ray job submit --no-wait --runtime-env="${RUNTIME_ENV}" \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.fsdp_config.param_offload=${offload} \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=-1 \

recipe/dapo/run_dapo_wo_ds_qwen2.5_32b.sh

@@ -105,6 +105,7 @@ ray job submit --no-wait --runtime-env="${RUNTIME_ENV}" \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.fsdp_config.param_offload=${offload} \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=-1 \

recipe/dapo/test_dapo_7b.sh

@@ -85,6 +85,7 @@ ray job submit --no-wait --runtime-env="${RUNTIME_ENV}" \
     actor_rollout_ref.rollout.log_prob_max_token_len_per_gpu=$((max_prompt_length + max_response_length)) \
     actor_rollout_ref.model.path="${MODEL_PATH}" \
     actor_rollout_ref.model.enable_gradient_checkpointing=True \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.actor.optim.lr=1e-6 \
     actor_rollout_ref.actor.optim.lr_warmup_steps=10 \
     actor_rollout_ref.actor.optim.weight_decay=0.1 \

recipe/dapo/test_dapo_7b_math.sh

@@ -82,6 +82,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.actor.ppo_max_token_len_per_gpu=${actor_ppo_max_token_len} \
     actor_rollout_ref.ref.log_prob_max_token_len_per_gpu=${infer_ppo_max_token_len} \
     actor_rollout_ref.rollout.log_prob_max_token_len_per_gpu=${infer_ppo_max_token_len} \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.model.path="${MODEL_PATH}" \
     actor_rollout_ref.model.enable_gradient_checkpointing=True \
     actor_rollout_ref.actor.optim.lr=1e-6 \

recipe/dapo/test_dapo_7b_math_lora.sh

@@ -81,6 +81,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.actor.ppo_max_token_len_per_gpu=${actor_ppo_max_token_len} \
     actor_rollout_ref.ref.log_prob_max_token_len_per_gpu=${infer_ppo_max_token_len} \
     actor_rollout_ref.rollout.log_prob_max_token_len_per_gpu=${infer_ppo_max_token_len} \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.model.path="${MODEL_PATH}" \
     actor_rollout_ref.model.enable_gradient_checkpointing=True \
     actor_rollout_ref.model.lora_rank=8 \

recipe/dapo/test_dapo_7b_math_megatron.sh

@@ -108,6 +108,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.megatron.pipeline_model_parallel_size=${train_pp} \
     actor_rollout_ref.ref.megatron.tensor_model_parallel_size=${train_tp} \
     actor_rollout_ref.ref.megatron.param_offload=${offload} \

recipe/dapo/test_dapo_dspk_671b_megatron.sh

@@ -115,6 +115,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.megatron.pipeline_model_parallel_size=${train_pp} \
     actor_rollout_ref.ref.megatron.tensor_model_parallel_size=${train_tp} \
     actor_rollout_ref.ref.megatron.expert_model_parallel_size=${train_ep} \

recipe/dapo/test_dapo_qwen3_30b_math.sh

@@ -102,6 +102,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.fsdp_config.param_offload=${offload} \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=${fsdp_size} \

recipe/dapo/test_dapo_qwen3_30b_math_single_node.sh

@@ -102,6 +102,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.fsdp_config.param_offload=${offload} \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=${fsdp_size} \

recipe/entropy/32b_clip_cov.sh

@@ -85,6 +85,7 @@ HYDRA_FULL_ERROR=1 python -m recipe.entropy.main_entropy \
     actor_rollout_ref.actor.policy_loss.clip_cov_ub=${clip_cov_ub} \
     actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=8 \
     actor_rollout_ref.rollout.mode=sync \
+    actor_rollout_ref.rollout.name=vllm \
     algorithm.adv_estimator=${adv_estimator} \
     algorithm.use_kl_in_reward=${use_kl_in_reward} \
     algorithm.kl_ctrl.kl_coef=${kl_coef} \

recipe/entropy/32b_kl_cov.sh

@@ -82,6 +82,7 @@ HYDRA_FULL_ERROR=1 python -m recipe.entropy.main_entropy \
     actor_rollout_ref.actor.policy_loss.ppo_kl_coef=${ppo_kl_coef} \
     actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=8 \
     actor_rollout_ref.rollout.mode=sync \
+    actor_rollout_ref.rollout.name=vllm \
     algorithm.adv_estimator=${adv_estimator} \
     algorithm.use_kl_in_reward=${use_kl_in_reward} \
     algorithm.kl_ctrl.kl_coef=${kl_coef} \

recipe/entropy/32b_kl_cov_mininbsz.sh

@@ -81,6 +81,7 @@ HYDRA_FULL_ERROR=1 python -m recipe.entropy.main_entropy \
     actor_rollout_ref.actor.policy_loss.ppo_kl_coef=${ppo_kl_coef} \
     actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=8 \
     actor_rollout_ref.rollout.mode=sync \
+    actor_rollout_ref.rollout.name=vllm \
     algorithm.adv_estimator=${adv_estimator} \
     algorithm.use_kl_in_reward=${use_kl_in_reward} \
     algorithm.kl_ctrl.kl_coef=${kl_coef} \

recipe/entropy/7b_clip_cov.sh

@@ -85,6 +85,7 @@ HYDRA_FULL_ERROR=1 python -m recipe.entropy.main_entropy \
     actor_rollout_ref.actor.policy_loss.clip_cov_ub=${clip_cov_ub} \
     actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=8 \
     actor_rollout_ref.rollout.mode=sync \
+    actor_rollout_ref.rollout.name=vllm \
     algorithm.adv_estimator=${adv_estimator} \
     algorithm.use_kl_in_reward=${use_kl_in_reward} \
     algorithm.kl_ctrl.kl_coef=${kl_coef} \

recipe/entropy/7b_kl_cov.sh

@@ -81,6 +81,7 @@ HYDRA_FULL_ERROR=1 python -m recipe.entropy.main_entropy \
     actor_rollout_ref.actor.policy_loss.ppo_kl_coef=${ppo_kl_coef} \
     actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=8 \
     actor_rollout_ref.rollout.mode=sync \
+    actor_rollout_ref.rollout.name=vllm \
     algorithm.adv_estimator=${adv_estimator} \
     algorithm.use_kl_in_reward=${use_kl_in_reward} \
     algorithm.kl_ctrl.kl_coef=${kl_coef} \

recipe/one_step_off_policy/dapo_7b_math_fsdp2_4_12.sh

@@ -112,6 +112,7 @@ python3 -m recipe.one_step_off_policy.main_ppo \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.fsdp_config.param_offload=${ref_offload} \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=${fsdp_size} \

recipe/one_step_off_policy/dapo_7b_math_fsdp2_colocate.sh

@@ -107,6 +107,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.fsdp_config.param_offload=${offload} \
     actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
     actor_rollout_ref.actor.fsdp_config.fsdp_size=${fsdp_size} \

recipe/one_step_off_policy/dapo_7b_math_megatron_4_12.sh

@@ -119,6 +119,7 @@ python3 -m recipe.one_step_off_policy.main_ppo \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.megatron.pipeline_model_parallel_size=${train_pp} \
     actor_rollout_ref.ref.megatron.tensor_model_parallel_size=${train_tp} \
     actor_rollout_ref.ref.megatron.param_offload=${ref_offload} \

recipe/one_step_off_policy/dapo_7b_math_megatron_colocate.sh

@@ -113,6 +113,7 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
     actor_rollout_ref.rollout.val_kwargs.do_sample=True \
     actor_rollout_ref.rollout.val_kwargs.n=1 \
+    actor_rollout_ref.rollout.name=vllm \
     actor_rollout_ref.ref.megatron.pipeline_model_parallel_size=${train_pp} \
     actor_rollout_ref.ref.megatron.tensor_model_parallel_size=${train_tp} \
     actor_rollout_ref.ref.megatron.param_offload=${offload} \

tests/special_e2e/run_one_step_off_policy.sh

@@ -89,6 +89,7 @@ common_params=(
     actor_rollout_ref.rollout.val_kwargs.do_sample=True
     actor_rollout_ref.rollout.val_kwargs.n=1
     actor_rollout_ref.rollout.enable_chunked_prefill=True \
+    actor_rollout_ref.rollout.name=vllm \
     reward_model.reward_manager=dapo
     +reward_model.reward_kwargs.overlong_buffer_cfg.enable=${enable_overlong_buffer}
     +reward_model.reward_kwargs.overlong_buffer_cfg.len=${overlong_buffer_len}

verl/trainer/config/_generated_ppo_megatron_trainer.yaml

@@ -121,7 +121,7 @@ actor_rollout_ref:
       save_path: null
     load_weight: true
   rollout:
-    name: vllm
+    name: ???
     mode: sync
     temperature: 1.0
     top_k: -1

verl/trainer/config/_generated_ppo_trainer.yaml

@@ -84,7 +84,7 @@ actor_rollout_ref:
     entropy_from_logits_with_chunking: false
     entropy_checkpointing: false
   rollout:
-    name: vllm
+    name: ???
     mode: sync
     temperature: 1.0
     top_k: -1

verl/trainer/config/rollout/rollout.yaml

@@ -1,5 +1,5 @@
 # actor_rollout_ref.rollout.name: hf/vllm/sglang. The default value will be removed in the future
-name: vllm
+name: ???
 
 # sync: LLM, async: AsyncLLM
 mode: sync