[misc] fix: validation batch repeat before feed into rollout (#614)

verl/trainer/ppo/ray_trainer.py

@@ -505,6 +505,10 @@ class RayPPOTrainer(object):
         for test_data in self.val_dataloader:
             test_batch = DataProto.from_single_dict(test_data)
 
+            # repeat test batch
+            test_batch = test_batch.repeat(repeat_times=self.config.actor_rollout_ref.rollout.val_kwargs.n,
+                                           interleave=True)
+
             # we only do validation on rule-based rm
             if self.config.reward_model.enable and test_batch[0].non_tensor_batch['reward_model']['style'] == 'model':
                 return {}
@@ -537,6 +541,7 @@ class RayPPOTrainer(object):
             # pad to be divisible by dp_size
             test_gen_batch_padded, pad_size = pad_dataproto_to_divisor(test_gen_batch, self.actor_rollout_wg.world_size)
             test_output_gen_batch_padded = self.actor_rollout_wg.generate_sequences(test_gen_batch_padded)
+
             # unpad
             test_output_gen_batch = unpad_dataproto(test_output_gen_batch_padded, pad_size=pad_size)
             print('validation generation end')

verl/workers/rollout/vllm_rollout/vllm_rollout.py

@@ -191,7 +191,7 @@ class vLLMRollout(BaseRollout):
                 'top_k': self.config.val_kwargs.top_k,
                 'top_p': self.config.val_kwargs.top_p,
                 'temperature': self.config.val_kwargs.temperature,
-                'n': self.config.val_kwargs.n,
+                'n': 1,  # if validate, already repeat in ray_trainer
             }
 
         # users can customize different sampling_params at different run

verl/workers/rollout/vllm_rollout/vllm_rollout_spmd.py

@@ -204,7 +204,7 @@ class vLLMRollout(BaseRollout):
                 'top_k': self.config.val_kwargs.top_k,
                 'top_p': self.config.val_kwargs.top_p,
                 'temperature': self.config.val_kwargs.temperature,
-                'n': self.config.val_kwargs.n,
+                'n': 1,  # if validate, already repeat in ray_trainer
             }
 
         # users can customize different sampling_params at different run