diff --git a/vllm/v1/attention/backends/mla/indexer.py b/vllm/v1/attention/backends/mla/indexer.py index ded3218346..5d12d27e76 100644 --- a/vllm/v1/attention/backends/mla/indexer.py +++ b/vllm/v1/attention/backends/mla/indexer.py @@ -303,7 +303,10 @@ class DeepseekV32IndexerMetadataBuilder(AttentionMetadataBuilder): device=self.device, ) self.arange_buffer = torch.arange( - scheduler_config.max_num_seqs * next_n, + max( + scheduler_config.max_num_seqs * next_n, + scheduler_config.max_num_batched_tokens, + ), dtype=torch.int32, device=self.device, )