From 716950be87c8f94c09760ab39216f4a002f36f71 Mon Sep 17 00:00:00 2001 From: weijie <609587797@qq.com> Date: Sat, 9 May 2026 19:35:58 +0800 Subject: [PATCH] Fix: skip trainer config validation in bench mode Bench mode runs explorer-only; cluster.trainer_gpu_num is left at 0 because the cluster validator (line 244) skips trainer GPU allocation for bench/explore/serve. The trainer config validator however still kept 'bench' in its whitelist, so any local-model bench run hit: trinity/trainer/verl/verl_config.py:430 if train_batch_size % (world_size // sp_size) != 0: ZeroDivisionError: integer division or modulo by zero Drop bench from the whitelist; same fast-path semantics as the existing external_model.enable check immediately below. --- trinity/common/config_validator.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/trinity/common/config_validator.py b/trinity/common/config_validator.py index 57069ad313f..729e5b2e97f 100644 --- a/trinity/common/config_validator.py +++ b/trinity/common/config_validator.py @@ -1165,7 +1165,7 @@ def validate(self, config: Config) -> None: or save checkpoint strategy is invalid. """ if ( - config.mode not in ["train", "both", "bench", "colocate"] + config.mode not in ["train", "both", "colocate"] and config.trainer.trainer_strategy != "megatron" ): return