[recipe] chore: use verl.utils.metric to import reduce_metrics (#3927)

HollowMan6 · web-flow · commit 52fc6747f4d5 · 2025-10-27T21:55:28.000+08:00
### What does this PR do? > Add **concise** overview of what this PR aims to achieve or accomplish. Reference related GitHub issues and PRs that help with the review. Fix FutureWarning: Warning: Function 'verl.trainer.ppo.metric_utils.reduce_metrics' is deprecated. Please use 'verl.utils.metric.reduce_metrics' instead. ### Checklist Before Starting - [X] Search for similar PRs. Paste at least one query link here: ... - [X] 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. - [X] Read the [Contribute Guide](https://github.com/volcengine/verl/blob/main/CONTRIBUTING.md). - [X] 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` - [X] Add / Update [the documentation](https://github.com/volcengine/verl/tree/main/docs). - [X] 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: ... - [X] 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).) Signed-off-by: Hollow Man <hollowman@opensuse.org>
diff --git a/recipe/dapo/dapo_ray_trainer.py b/recipe/dapo/dapo_ray_trainer.py
@@ -28,12 +28,7 @@
 
 from verl import DataProto
 from verl.trainer.ppo.core_algos import agg_loss
-from verl.trainer.ppo.metric_utils import (
-    compute_data_metrics,
-    compute_throughout_metrics,
-    compute_timing_metrics,
-    reduce_metrics,
-)
+from verl.trainer.ppo.metric_utils import compute_data_metrics, compute_throughout_metrics, compute_timing_metrics
 from verl.trainer.ppo.ray_trainer import (
     AdvantageEstimator,
     RayPPOTrainer,
@@ -42,6 +37,7 @@
     compute_response_mask,
 )
 from verl.trainer.ppo.reward import compute_reward
+from verl.utils.metric import reduce_metrics
 from verl.utils.profiler import marked_timer
 from verl.utils.rollout_skip import RolloutSkip
 
diff --git a/recipe/entropy/entropy_ray_trainer.py b/recipe/entropy/entropy_ray_trainer.py
@@ -26,12 +26,7 @@
 from tqdm import tqdm
 
 from verl import DataProto
-from verl.trainer.ppo.metric_utils import (
-    compute_data_metrics,
-    compute_throughout_metrics,
-    compute_timing_metrics,
-    reduce_metrics,
-)
+from verl.trainer.ppo.metric_utils import compute_data_metrics, compute_throughout_metrics, compute_timing_metrics
 from verl.trainer.ppo.ray_trainer import (
     AdvantageEstimator,
     RayPPOTrainer,
@@ -40,6 +35,7 @@
     compute_response_mask,
 )
 from verl.trainer.ppo.reward import compute_reward
+from verl.utils.metric import reduce_metrics
 from verl.utils.profiler import simple_timer
 
 
diff --git a/recipe/spin/spin_trainer.py b/recipe/spin/spin_trainer.py
@@ -36,14 +36,10 @@
 from verl.protocol import pad_dataproto_to_divisor, unpad_dataproto
 from verl.single_controller.ray import RayClassWithInitArgs, RayResourcePool, RayWorkerGroup
 from verl.single_controller.ray.base import create_colocated_worker_cls
-from verl.trainer.ppo.metric_utils import (
-    compute_throughout_metrics,
-    compute_timing_metrics,
-    process_validation_metrics,
-    reduce_metrics,
-)
+from verl.trainer.ppo.metric_utils import compute_throughout_metrics, compute_timing_metrics, process_validation_metrics
 from verl.trainer.ppo.utils import Role, WorkerType, need_reference_policy, need_reward_model
 from verl.utils.checkpoint.checkpoint_manager import find_latest_ckpt_path
+from verl.utils.metric import reduce_metrics
 from verl.utils.seqlen_balancing import get_seqlen_balanced_partitions, log_seqlen_unbalance
 from verl.utils.torch_functional import masked_mean
 from verl.utils.tracking import ValidationGenerationsLogger
diff --git a/recipe/sppo/sppo_ray_trainer.py b/recipe/sppo/sppo_ray_trainer.py
@@ -33,7 +33,6 @@
 from verl.single_controller.ray import RayWorkerGroup
 from verl.trainer.ppo import core_algos
 from verl.trainer.ppo.core_algos import agg_loss
-from verl.trainer.ppo.metric_utils import reduce_metrics
 from verl.trainer.ppo.ray_trainer import (
     AdvantageEstimator,
     RayPPOTrainer,
@@ -43,6 +42,7 @@
 )
 from verl.trainer.ppo.reward import compute_reward, compute_reward_async
 from verl.trainer.ppo.utils import Role, WorkerType, need_reference_policy, need_reward_model
+from verl.utils.metric import reduce_metrics
 from verl.utils.profiler.performance import simple_timer
 from verl.utils.tracking import ValidationGenerationsLogger
 
