fix: handle None multimodal_inputs during merging and filtering batches in...

fix: handle None multimodal_inputs during merging and filtering batches in disaggregation decode mode (#6169)

python/sglang/srt/managers/schedule_batch.py

@@ -1485,6 +1485,7 @@ class ScheduleBatch(ScheduleBatchDisaggregationDecodeMixin):
             self.encoder_lens_cpu = [self.encoder_lens_cpu[i] for i in keep_indices]
 
         self.reqs = [self.reqs[i] for i in keep_indices]
+        if self.multimodal_inputs is not None:
             self.multimodal_inputs = [self.multimodal_inputs[i] for i in keep_indices]
         self.req_pool_indices = self.req_pool_indices[keep_indices_device]
         self.seq_lens = self.seq_lens[keep_indices_device]
@@ -1534,6 +1535,7 @@ class ScheduleBatch(ScheduleBatchDisaggregationDecodeMixin):
             self.top_logprobs_nums = [0] * len(self.reqs) + other.top_logprobs_nums
             self.token_ids_logprobs = [None] * len(self.reqs) + other.token_ids_logprobs
         self.reqs.extend(other.reqs)
+        if self.multimodal_inputs is not None:
             self.multimodal_inputs.extend(other.multimodal_inputs)
 
         self.return_logprob |= other.return_logprob